Intelligent agent in communication network
By determining the need for intelligent agents and orchestrating their collaboration, the solution enhances network performance and automation by enabling efficient execution of complex tasks in communication networks.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- NOKIA SOLUTIONS (SHANGHAI) CO LTD
- Filing Date
- 2024-12-19
- Publication Date
- 2026-06-25
AI Technical Summary
Existing communication networks lack efficient methods for orchestrating intelligent agents to collaborate and optimize network performance and automation, limiting the full potential of intelligent agents in complex tasks.
A first apparatus determines the need for an intelligent agent (IA) to respond to a request, decomposes the request into an execution plan, and transmits the plan to a second apparatus hosting a target IA, enabling collaboration among different intelligent agents to enhance network performance and automation.
This approach improves network performance and automation by leveraging the full potential of intelligent agents through coordinated task execution and collaboration, addressing the limitations of existing systems.
Smart Images

Figure CN2024140770_25062026_PF_FP_ABST
Abstract
Description
INTELLIGENT AGENT IN COMMUNICATION NETWORKFIELD
[0001] Various example embodiments of the present disclosure generally relate to the field of telecommunication and in particular, to methods, devices, apparatuses and computer readable storage medium for an intelligent agent (IA) in a communication network.BACKGROUND
[0002] A communication network may serve as a facility that enables communications between two or more communication devices or provides communication devices access to a data network. A mobile or wireless communication network is one example of a communication network. A communication device may be provided with a service by an application server.
[0003] The communication network may operate in accordance with standards such as those provided by Third Generation Partnership Project (3GPP) or European Telecommunications Standards Institute (ETSI) . Examples of standards provided by 3GPP are the so-called 3GPP standards for cellular technology generations, such as 3GPP standards for 4G technology, 5G technology, 6G technology etc.SUMMARY
[0004] In a first aspect of the present disclosure, there is provided a first apparatus. The first apparatus comprises at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the first apparatus to: in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determine, based on the request, an execution plan to be executed by at least one IA; and transmit an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.
[0005] In a second aspect of the present disclosure, there is provided a method. The method comprises: in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determining based on the request, an execution plan to be executed by at least one IA; and transmitting an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.
[0006] In a third aspect of the present disclosure, there is provided a first apparatus. The first apparatus comprises means for in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determining based on the request, an execution plan to be executed by at least one IA; and means for transmitting an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.
[0007] In a fourth aspect of the present disclosure, there is provided a computer readable medium. The computer readable medium comprises instructions stored thereon for causing an apparatus to perform at least the method according to the second aspect.
[0008] It is to be understood that the Summary section is not intended to identify key or essential features of embodiments of the present disclosure, nor is it intended to be used to limit the scope of the present disclosure. Other features of the present disclosure will become easily comprehensible through the following description.BRIEF DESCRIPTION OF THE DRAWINGS
[0009] Some example embodiments will now be described with reference to the accompanying drawings, where:
[0010] FIG. 1 illustrates an example communication environment in which example embodiments of the present disclosure can be implemented;
[0011] FIG. 2 illustrates a signaling flow of an example process for IA orchestration in accordance with some example embodiments of the present disclosure;
[0012] FIG. 3 illustrates a schematic diagram of an intelligent agent framework in accordance with some example embodiments of the present disclosure;
[0013] FIG. 4 illustrates an example Unified Modeling Language (UML) diagram of an information model for IA orchestration according to some example embodiments of the present disclosure;
[0014] FIG. 5 illustrates a signaling flow of an example between a consumer and a producer concerning the lifecycle of a task or intent in accordance with some example embodiments of the present disclosure;
[0015] FIG. 6 illustrates a signaling flow of an example procedure for agent discovery and configuration in accordance with some example embodiments of the present disclosure;
[0016] FIG. 7 illustrates an example diagram of a use case example where intelligent agents are aiding network management and optimization in accordance with some example embodiments of the present disclosure;
[0017] FIG. 8 illustrates an example diagram of a use case example where intelligent agents are tailored to consumer-oriented tasks in accordance with some example embodiments of the present disclosure;
[0018] FIG. 9 illustrates a flowchart of a method implemented at a first apparatus in accordance with some example embodiments of the present disclosure;
[0019] FIG. 10 illustrates a simplified block diagram of a device that is suitable for implementing example embodiments of the present disclosure; and
[0020] FIG. 11 illustrates a block diagram of an example computer readable medium in accordance with some example embodiments of the present disclosure.
[0021] Throughout the drawings, the same or similar reference numerals represent the same or similar element.DETAILED DESCRIPTION
[0022] Principle of the present disclosure will now be described with reference to some example embodiments. It is to be understood that these embodiments are described only for the purpose of illustration and help those skilled in the art to understand and implement the present disclosure, without suggesting any limitation as to the scope of the disclosure. Embodiments described herein can be implemented in various manners other than the ones described below.
[0023] In the following description and claims, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure belongs.
[0024] References in the present disclosure to “one embodiment, ” “an embodiment, ” “an example embodiment, ” and the like indicate that the embodiment described may include a particular feature, structure, or characteristic, but it is not necessary that every embodiment includes the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to affect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.
[0025] It shall be understood that although the terms “first, ” “second, ” …, etc. in front of noun (s) and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another and they do not limit the order of the noun (s) . For example, a first element could be termed a second element, and similarly, a second element could be termed a first element, without departing from the scope of example embodiments. As used herein, the term “and / or” includes any and all combinations of one or more of the listed terms.
[0026] As used herein, “at least one of the following: <a list of two or more elements>” and “at least one of <a list of two or more elements>” and similar wording, where the list of two or more elements are joined by “and” or “or” , mean at least any one of the elements, or at least any two or more of the elements, or at least all the elements.
[0027] As used herein, unless stated explicitly, performing a step “in response to A” does not indicate that the step is performed immediately after “A” occurs and one or more intervening steps may be included.
[0028] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a” , “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” , “comprising” , “has” , “having” , “includes” and / or “including” , when used herein, specify the presence of stated features, elements, and / or components etc., but do not preclude the presence or addition of one or more other features, elements, components and / or combinations thereof.
[0029] As used in this application, the term “circuitry” may refer to one or more or all of the following: (a) hardware-only circuit implementations (such as implementations in only analog and / or digital circuitry) and (b) combinations of hardware circuits and software, such as (as applicable) : (i) a combination of analog and / or digital hardware circuit (s) with software / firmware and (ii) any portions of hardware processor (s) with software (including digital signal processor (s) ) , software, and memory (ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions) and (c) hardware circuit (s) and or processor (s) , such as a microprocessor (s) or a portion of a microprocessor (s) , that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation.
[0030] This definition of circuitry applies to all uses of this term in this application, including in any claims. As a further example, as used in this application, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.
[0031] As used herein, the term “communication network” refers to a network following any suitable communication standards, such as New Radio (NR) , Long Term Evolution (LTE) , LTE-Advanced (LTE-A) , Wideband Code Division Multiple Access (WCDMA) , High-Speed Packet Access (HSPA) , Narrow Band Internet of Things (NB-IoT) and so on. Furthermore, the communications between a terminal device and a network device in the communication network may be performed according to any suitable generation communication protocols, including, but not limited to, the first generation (1G) , the second generation (2G) , 2.5G, 2.75G, the third generation (3G) , the fourth generation (4G) , 4.5G, the fifth generation (5G) , 5.5G, the sixth generation (6G) communication protocols, and / or any other protocols either currently known or to be developed in the future. Embodiments of the present disclosure may be applied in various communication systems. Given the rapid development in communications, there will of course also be future type communication technologies and systems with which the present disclosure may be embodied. It should not be seen as limiting the scope of the present disclosure to only the aforementioned system.
[0032] As used herein, the term “network device” refers to a node in a communication network via which a terminal device accesses the network and receives services therefrom. The network device may refer to a base station (BS) or an access point (AP) , for example, a node B (NodeB or NB) , an evolved NodeB (eNodeB or eNB) , an NR NB (also referred to as a gNB) , a Remote Radio Unit (RRU) , a radio header (RH) , a remote radio head (RRH) , a relay, an Integrated Access and Backhaul (IAB) node, a low power node such as a femto, a pico, a non-terrestrial network (NTN) or non-ground network device such as a satellite network device, a low earth orbit (LEO) satellite and a geosynchronous earth orbit (GEO) satellite, an aircraft network device, and so forth, depending on the applied terminology and technology. In some example embodiments, radio access network (RAN) split architecture comprises a Centralized Unit (CU) and a Distributed Unit (DU) at an IAB donor node. An IAB node comprises a Mobile Terminal (IAB-MT) part that behaves like a UE toward the parent node, and a DU part of an IAB node behaves like a base station toward the next-hop IAB node.
[0033] The term “terminal device” refers to any end device that may be capable of wireless communication. By way of example rather than limitation, a terminal device may also be referred to as a communication device, user equipment (UE) , a Subscriber Station (SS) , a Portable Subscriber Station, a Mobile Station (MS) , or an Access Terminal (AT) . The terminal device may include, but not limited to, a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones, a tablet, a wearable terminal device, a personal digital assistant (PDA) , portable computers, desktop computer, image capture terminal devices such as digital cameras, gaming terminal devices, music storage and playback appliances, vehicle-mounted wireless terminal devices, wireless endpoints, mobile stations, laptop-embedded equipment (LEE) , laptop-mounted equipment (LME) , USB dongles, smart devices, wireless customer-premises equipment (CPE) , an Internet of Things (IoT) device, a watch or other wearable, a head-mounted display (HMD) , a vehicle, a drone, a medical device and applications (e.g., remote surgery) , an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated processing chain contexts) , a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like. The terminal device may also correspond to a Mobile Termination (MT) part of an IAB node (e.g., a relay node) . In the following description, the terms “terminal device” , “communication device” , “terminal” , “user equipment” and “UE” may be used interchangeably.
[0034] As used herein, the term “resource, ” “transmission resource, ” “resource block, ” “physical resource block” (PRB) , “uplink resource, ” or “downlink resource” may refer to any resource for performing a communication, for example, a communication between a terminal device and a network device, such as a resource in time domain, a resource in frequency domain, a resource in space domain, a resource in code domain, or any other combination of the time, frequency, space and / or code domain resource enabling a communication, and the like. In the following, unless explicitly stated, a resource in both frequency domain and time domain will be used as an example of a transmission resource for describing some example embodiments of the present disclosure. It is noted that example embodiments of the present disclosure are equally applicable to other resources in other domains.
[0035] As used herein, the term “intelligent agent” (IA) (also referred to as agent in short) may refer to an entity that makes decisions and take actions based on perception of its environment. The IA may be implemented based on any suitable algorithm for example, based on rules, classical machine learning models, or deep learning models, etc. The intelligent agent may act as an autonomous system capable of perceiving its environment using different data modalities and leveraging on different tools in order to derive decisions and take actions. In telecommunication domain, the agent may correspond to a product or solution capable of performing specific telco task, e.g., adjusting the power level of gNB.
[0036] Generative artificial intelligence (AI) is an approach to create a new content of different types following the characteristics of the training data. One of the approaches of generative AI is large language models (LLMs) used to generate plausible text / language based on the input query. There are currently numerous examples of proprietary and open-source models available and used in different applications. The generic LLM (foundation model) are usually obtained by extensive training using huge amount of data in order to capture the relations between the words and obtain generic capabilities for text understanding, processing, and generation. Such process is called pre-training. The fine-tuning is a process of adapting a generic model towards domain specific tasks, such as understanding technical text and recommending management actions in telco domains. This may be done by selectively adjusting / training a subset of model parameters or a set of newly added parameters.
[0037] LLMs use one input and output data modality and that is language. Large Multimodal Models (LMM) combine various data modalities, e.g. text, audio, visual, sensor data etc. capturing the correlations between different modalities. Such approach may be applicable to any kind of data, incl. network data. Finally, Small Language Models (SLM) are less compute intense than LLMs, both in training and inference, with good performance especially if trained and used for specific problem, therefore besides LLMs and LMMs have also high relevance in telco applications.
[0038] The AI agents are commonly defined as entities that make decisions and take actions based on perception of their environment. There may be different types of AI agents, e.g. those based on rules, based on reinforcement learning, LLM / LMM / SLM-based etc. LLM / LMM / SLM-based agents leverage on LLM / LMM / SLMs serving as the “brain” expanding their ability to perceive the environment and to take actions by using strategies such as external tools (e.g. internet search) calling or multimodal perception of environment considering input data of different types / modalities, e.g. audio, visual, text, network KPIs and metrics etc.
[0039] Chain-of-Thought (CoT) is another strategy that enables reasoning and planning by making the LLM providing the output along with step-by-step description, i.e., “chain of thought” while splitting the high-level task into smaller tasks. The LLM-based agents show performance improvements due to their capability to capture knowledge, interpret instructions, reason, etc. In addition, the LLM based agents can understand the language and multimodal inputs, such as customer documentation, ticket resolution reports, user feedback, etc.
[0040] LLMs have showed significant reasoning capabilities. In addition, LLMs could be augmented with other components such as memory, utilization of different tools, environment perception, and critics or grounding, i.e. output generation based on the relevant information e.g. use-case specific, not available as part of training data, leading to development of LLM agents interacting with environment and performing autonomous actions.
[0041] Such agents may interact with each other in order to complete the high-level or complex task. In such a multi-agent system, the agents’ collaboration is crucial for completing the high-level task via completion of set of smaller tasks, e.g. search, optimization, resource allocation etc.
[0042] An agent normally has a few components: Memory, Planning, Action Execution. It may additionally have a policy component or critics. In the network, each functional entity may have at least one agent supported by at least one LLM / LMM / SLM.
[0043] Each agent may perceive the network state between the functional entities themselves and those directly related to them, via e.g., use tools / API for external related network environments data, consumer service from other Network functions to fetch e.g., FCAP data and perform analytics for the collected up-to-date data for network state analytics, or use the capabilities from other agents in the related network entities.
[0044] Each agent may process and save the collected, generated analytics data and other job related context into memory, which may be retrieved for the whole lifecycle of request execution. Each agent may decompose the request from consumer (e.g., user, other Agents) to tasks or action plans and execute the action plan, then generate the final results. Each agent may have a policy control for planning, action execution, and working memory capabilities.
[0045] In the case of intelligent agents, enabling the controlled utilization of agents in telecommunication systems as well as enabling and controlling their collaborations is fundamental for unlocking full potential of intelligent agents.
[0046] In accordance with some example embodiments of the present disclosure, there is provided a solution for an IA in a communication network. In the solution, a first apparatus which obtains a request determines whether an IA is needed for responding to the request. If the IA is needed, the first apparatus determines, based on the request, an execution plan to be executed by at least one IA. Then, the first apparatus transmits an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.
[0047] Instead of a specific individual IA targeting one (simple) , with the proposed solution, mutual collaboration among different intelligent agents may be achieved. In this way, task adds to improvement of network performance and automation, the full potential of intelligent agents may be exploited to solve complex tasks. As such, network performance and automation may be improved.
[0048] Example embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.
[0049] Reference is first made to FIG. 1. FIG. 1 illustrates an example communication environment 100 in which example embodiments of the present disclosure may be implemented. The communication environment 100 may comprise a first apparatus 110 and a second apparatus 120 in communication. The second apparatus 120 hosts an intelligent agent 130. The first apparatus 110 and the second apparatus 120 may be any suitable apparatus in the communication network. For example, the first apparatus 110 may be or may be comprised in a terminal device, a radio access network (RAN) device, a core network (CN) function / element, a management device / function, or a third-party application / service. Similarly, the second apparatus 120 may be or may be comprised in a terminal device, a radio access network (RAN) device, a core network (CN) function / element a management device / function, or a third-party application / service.
[0050] Communications in the communication environment 100 may be implemented according to any proper communication protocol (s) , comprising, but not limited to, cellular communication protocols, wireless local network communication protocols such as Institute for Electrical and Electronics Engineers (IEEE) 802.11 and the like, and / or any other protocols currently known or to be developed in the future. Moreover, the communication may utilize any proper wireless communication technology, comprising 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 (OFDM) , Discrete Fourier Transform spread OFDM (DFT-s-OFDM) and / or any other technologies currently known or to be developed in the future.
[0051] It is to be understood that the number of apparatuses and their connection shown in FIG. 1 are only for the purpose of illustration without suggesting any limitation. The communication environment 100 may include any suitable number of apparatuses configured to implement example embodiments of the present disclosure. Although not shown, it would be appreciated that one or more additional apparatuses may be deployed in the communication environment 100. For example, the communication environment 100 may comprise further apparatuses in communication with the first apparatus 110, e.g., a third apparatus that transmits a request to the first apparatus 110.
[0052] FIG. 2 illustrates a signaling flow of an example process 200 for IA orchestration in accordance with some example embodiments of the present disclosure. For the purpose of discussion, the process 200 will be described with reference to the first apparatus 110 and the second apparatus 120 in FIG. 1.
[0053] In the following, the first apparatus 110 may be also referred to as the IA orchestration apparatus configured to perform IA orchestration or management. The IA may be based on any suitable artificial intelligence / machine learning (AI / ML) model, e.g., an LLM, an LMM, or a SLM. The scope of the present disclosure is not limited in this regard.
[0054] In some embodiments, the first apparatus 110 may be a newly defined core network function named intelligent agent orchestration function (IAOF) . Alternatively, the first apparatus 110 may be implemented in an existing core network function, e.g., AI / ML inference function. In some embodiments, the first apparatus 110 may be implemented in a terminal device, a management node / function, a RAN node, or a third party device or application.
[0055] As illustrated in FIG. 2, in the process 200, in accordance with a determination that an IA is needed for responding to a request, the first apparatus 110 determines 212, based on the request, an execution plan to be executed by at least one IA. The request may be received by the first apparatus 110 from a further apparatus or generated from the first apparatus 110 itself. The first apparatus 110 further transmits 215 an execution task of the execution plan to the second apparatus 120 hosting a target IA of the at least one IA. The second apparatus 120 (i.e., the target IA) receives 218 the execution task and may execute (not shown) the received execution task.
[0056] The execution plan may refer to a list of actions, actionable (sub) tasks or execution (sub) tasks with some certain dependency and / or constraints. The action or (sub) task in the list may be executed directly or via an agent. Depending on the request, the execution plan may comprise various execution tasks. For example, if the request is an intent e.g. “I wish to get the RAN energy consumption in area X below Y kwh” , the execution plan may comprise specific tasks such as transmit power adjustment, cell switch on / off and selection of most energy efficient UPF. The specific tasks in the execution plan may be executed by respective IAs determined based on the request.
[0057] FIG. 3 illustrates a schematic diagram of an intelligent agent framework 300 in accordance with some example embodiments of the present disclosure. The intelligent agents are working together for a specific task, e.g. fulfilling network operator intent to reduce energy consumption by certain percentage. The involved intelligent agents may have no special relationship, or they may be structured, or organized in a hierarchical manner.
[0058] An intelligent agent 301 (called a planning agent) receives the task (called task A, e.g. network operator intent) from a consumer 330. The intelligent agent 301 needs to decompose the task A into smaller sub tasks (for example task N1, task N2, …, task Nm) . FIG. 3 shows task N1, task N2 and task N3 as an example. Each subtask may be received by an intelligent agent. For example, the task N1, task N2 and task N3 are received by the intelligent agent 311, the intelligent agent 312, and the intelligent agent 313, respectively. The task Nm may be further decomposed by another receiving intelligent agent (for example, agent N1 …agent Nn) , however this level of decomposition is not visible to intelligent agent A. The assumption is that the sub task intelligent agents will be domain specific backed by a domain specific LLM / LMM / SLM. For example, the agent 311 may perform gNB power level adjustments, for which the executor 321 may be a gNB 1 as shown in FIG. 3. The agent 312 may perform energy aware UPF selection in core network, for which the executor 322 may be an SMF or UPF. The agent 313 may perform any action, for which the executor 322 may be a gNB N. It is noted that the actions described with respect to FIG. 3 are examples without any limitation. The challenge in the scenario as shown in FIG. 3 is how the intelligent agent 301 can know which intelligent agents are performing which tasks, what are their capabilities and requirements. The intelligent agent 301 may also monitor and control the completion of the task A by tracking the progress of all involved intelligent agents and evaluating the results of intelligent agent Nx actions. Some example embodiments addressing these challenges will be described below.
[0059] In some embodiments, based on the request, the first apparatus 110 may first determine whether an IA is needed for responding to the request. If the first apparatus 110 determines that an IA is needed, the first apparatus 110 may determine the execution plan and the at least one IA to execute the execution plan. In some embodiments, the first apparatus 110 may determine whether the request can map to a set of create, read, update, delete (CRUD) operations. If so, the first apparatus 110 may determine that the request may be responded by the simple CRUD operations such that an IA is not needed. On the contrary, if the first apparatus 110 determines that the request cannot map to the simple CRUD operations, the first apparatus 110 may determine that an IA is needed for responding to the request.
[0060] In some embodiments, the first apparatus 110 may determine the at least one IA for executing the execution plan by using a first IA hosted in the first apparatus 110. The first IA hosted in the first apparatus 110 may be also referred to as a primary agent or planning agent in the following. The planning agent may select a list of IAs corresponding to the execution plan and / or configure the collaboration among the selected IAs.
[0061] In some embodiments, based on determining that an IA is needed for responding to the received request, the first apparatus 110 may instantiate the first IA for planning the execution plan, i.e., determining the execution plan and the corresponding IA (s) to execute the execution task (s) in the execution plan. Alternatively, the first IA hosted in the first apparatus 110 may be installed by the telecommunication system.
[0062] Alternatively, the first apparatus 110 may determine, based on the request, the at least one IA comprising the first IA hosted in the first apparatus. That is, the first apparatus 110 may use means other than an IA to determine the execution plan and then instantiate the first IA in the apparatus 110 to execute the execution plan. For example, a predetermined mapping of the request to the execution plan may be used.
[0063] In some embodiments, the first apparatus 110 may determine the execution plan to be executed by the at least one IA by decomposing the request into a plurality of execution tasks having an execution dependency and / or execution constraint. For example, the first apparatus 110 may decompose or break down the request to the execution plan considering task dependency, task constraints, agent dependency and / or agent constraints.
[0064] In some embodiments, the first apparatus 110 may determine the at least one IA to execute the execution plan based on discovery and / or registration of IAs (and their capabilities) . For example, a plurality of IAs may be registered with a core network element and their corresponding capability information may be stored. The first apparatus 110 may use any suitable means to discover the IAs and their IA capability information, and select from the discovered IAs, based on the request and the IA capability information, the at least one IA for executing the determined execution plan.
[0065] In some embodiments, the first apparatus 110 may receive the IA capability information from a third apparatus that transmits the request to the first apparatus. The third apparatus may be also referred to as a consumer which requests a service from the first apparatus 110 (also referred to as a producer in this case) . In some embodiments, the first apparatus 110 may receive the IA capability information from the consumer as at least part of IA provisioning information. The IA provisioning information may refer to provisioning / requirement or a configuration determined by the consumer for IA (s) used for responding to the request. For example, the IA capability information in the IA provisioning information may indicate a specific model type of an agent, and the first apparatus 110 may thus select a target IA supporting this specific model type to execute the execution plan for responding to the request. Further details of the IA provisioning information will be described in the following.
[0066] Alternatively or additionally, the first apparatus 110 may receive the IA capability information from a fourth apparatus responsible for an exposure or discovery service of the IA capability information, for example, a core network element the IAs are registered with. For example, the first apparatus 110 may receive the IA capability information in a discovery procedure.
[0067] In some embodiments, the IA capability information may comprise IA planning capability related to planning an execution plan as discussed above. For example, the first apparatus 110 may select a planning IA based on exposed planning capabilities of different IA in the network (and optional IA provisioning information transmitted from the consumer) .
[0068] Alternatively or additionally, the IA capability information may comprise information of at least one task supported by an agent, for example, a list of tasks the IA is capable to perform, or description of task (s) the IA is capable to perform, etc.
[0069] Alternatively or additionally, the IA capability information may comprise a capability of a base model for an agent. For example, the model-related capability may comprise a type of (generative) model used by the IA, e.g. LLM, LMM, etc. The model-related capability may comprise a size of the model powering the agent (e.g. large if the number of parameters exceeds 1 billion, small otherwise) . The model-related capability may comprise a training approach used to obtain the model powering the IA. For example, pre-training may be used to obtain a general-purpose model such that the corresponding IA may have generic capabilities such as identifying relationships between input data. Fine-tuning may be used to obtain a model specific to some domain or task, such that the corresponding IA may have specific capabilities to handle a specific task in a specific network domain. Internal knowledge capturing specific domain of operation, such as in RAN or core may be used for fine-tuning.
[0070] Alternatively or additionally, the IA capability information may comprise a memory / context length capability related to the length of memory / context the agent support. For example, an LLM model supports a limited memory length (context length) , and the memory length for an agent is not longer than the context length of the model.
[0071] Alternatively or additionally, the IA capability information may comprise a capability to use external knowledge (i.e., knowledge external to the model) or external tool, e.g. Internet search. Alternatively or additionally, the IA capability information may comprise a capability to collaborate with a further agent. For example, a standalone IA may be specified to perform a single task and not intended to receive inputs from other IA(s) . A cooperative IA may require inputs from other IA (s) .
[0072] Alternatively or additionally, the IA capability information may comprise a critic capability for verifying and refining an output from a further agent. For example, an IA may be able to verify output of other IA (s) and to provide recommendations on their refinements. Alternatively or additionally, the IA capability information may comprise a fallback capability. It may indicate the fallback option for addressing a task of the IA when the task failed.
[0073] Alternatively or additionally, the IA capability information may comprise an accountability and dependability capability. It may indicate e.g., IA availability in space (e.g., areas of Interest represented by geographical coordinates or a list of RAN nodes) , IA availability in time (e.g., time of day) , task accomplishment reliability (e.g., “in-time” task fulfillment) , and / or maintainability / fault recovery.
[0074] Alternatively or additionally, the IA capability information may comprise a data aspect of the IA. Examples may comprise an amount of data used for training the model powering the IA (e.g., a number of tokens, data size) and a type of data used for training and expected as input to the model. Examples of the type of data may comprise a language or a list of applicable languages. Examples of the type of data may further comprise mobile network data, e.g., network metrics, Key Performance Indicators (KPIs) , Fault, Configuration, Accounting, Performance (FCAP) Management data. Examples of the type of data may further comprise other related data such as data obtained from simulations, lab tests, filed test; customer specification documentation related to products and features; and historical records on network operations, such as troubleshooting ticket resolutions, service offering, customer satisfaction reports, etc. Examples of the type of data may further comprise audio data, video data, sensor data, etc.
[0075] In some embodiments, an information element (IE) may be used to indicate the IA capability information. Reference is now made to FIG. 4. FIG. 4 illustrates an example diagram of an information model 400 for IA orchestration according to some example embodiments of the present disclosure. The information model 400 includes information object classes (IOCs) and data types needed to realize the IA orchestration.
[0076] As shown in FIG. 4, a data type named as AgentCapability 404 may be used to indicate the IA capability information. The data type AgentCapability 404 may include the following attributes as shown in Table 1. In Table 1 and following tables, M indicates mandatory, O indicates optional, C indicates conditional. Table 1 is attributes of IA capability information
[0077] In Table 1, the attribute “supportedTaskTypeList” indicates the list of tasks that the agent is capable to perform along with task descriptions. The attribute “baseModelCapabilities” indicates the capabilities of the large model backed to the agent, or agent connected model type. These capabilities may include for example domain specific strength, internal knowledge capturing specific domain of operation, such as RAN or core; data aspects of the model (e.g., amount or type of data used for model training) ; model capability from the training, fine tuning (e.g., indicated as capability to solve generic or specific tasks) .
[0078] The attribute “useExternalKnowledge” indicates the capability of usage of external knowledge other than the Knowledge from model directly. The attribute “externalKnowledgeTypeList” indicates the type of external knowledge.
[0079] The attribute “collaborationWithOtherAgent” indicates the cooperation capability with another agent. The attribute “usingTool” indicates the use of external tools. The attribute “memoryLength” indicates the length of memory the agent support. The attribute “criticsCapabilities” indicates the capability to verify output of other intelligent agents and to provide recommendations on their refinements.
[0080] In some embodiments, the first apparatus 110 may receive the IA provisioning information, and determine the execution plan to be executed by the at least one IA based on the IA provisioning information. In some embodiments, the first apparatus 110 may further receive an update of the IA provisioning information during runtime of the execution plan. As discussed above, the IA provisioning information may comprise the IA capability information transmitted from the consumer.
[0081] In some embodiments, the IA provisioning information may further comprise IA agentic information. The agentic information may refer to general information of an IA. For example, the agentic information may comprise agent capability information as discussed above, an agent type, a potential collaboration agent or a list of potential collaboration agents or a condition for filtering collaboration agent (s) , an agent accountability score, and / or a policy for an agent. The agent accountability score may indicate accountability of the agent. It may be a float within a range of [0, 1] , and a higher value means higher accountability. It may be determined based on an evaluation from previous task execution by self-evaluation, feedback from the consumer or an overall evaluation. The policy for an agent may indicate a control policy to control agent capabilities and details of the provisioning policy will be described below.
[0082] The information model 400 shown in FIG. 4 includes an IOC “IntelligentAgent” 403 to indicate the agentic information. Table 2 shows example attributes of the agentic information. Table 2. Example attributes of agentic information
[0083] The attribute “agentType” indicates type of the agent. The attribute “agentCapabilities” with type as “AgentCapabilities” indicates abilities of the agent. The capability may be deriving from the pre-training, fine-tuning large models, or using of external knowledge. It may be expressed as: the list of tasks that the agent is capable to perform along with task descriptions; capabilities of the large model backed to the agent, or agent connected model type; usage of external knowledge, type of external knowledge; a cooperation capability with other agent (s) ; use of external tools; infrastructure capability (optional) ; and critics capabilities. The capabilities may include domain specific strength, internal knowledge capturing specific domain of operation, such as RAN or core; area (expressed in e.g., geographical coordinates or as a list of RAN nodes) of focus of the agent; data aspects of the model (e.g., amount or type of data used for model training) ; model capability from the training, fine tuning (e.g., indicated as capability to solve generic or specific tasks) .
[0084] The attribute “agentAccountbilityScore” indicates accountability of the agent. It may be a float with range 0 ...1, higher value means more accountable. It may be an evaluate from previous task execution by self-evaluation, feedback from consumer or an overall evaluation.
[0085] The attribute “collaborationAgentSelectionOptions” indicates conditions to filter the agents to collaborate with for subtasks. Example filters may include location or data modality filters.
[0086] The attribute “policyForAgent” (with type: “PolicyForAgent” ) indicates control policy to control agent capabilities. It may include: activate / deactivate agent’s capability; “ExecutionStrategy” indicating priority of task execution, bestPerformance, balanced, bestEnergySaving (hence less interaction with another Agent) ; “ScopeOfExternalInteraction” which is a filter about the scope per GeoLocation, per Network Function list, or AgentType, etc. ; “MaxNumberOfExternalAgent” indicating maximum number of external agent for a given Request; “fallbackCapabilities” indicating when a sub task request failed, the fallback option for addressing the sub-task.
[0087] In some example embodiments, the IA policy information may include at least one of: a managed activation / deactivation scope for agent capabilities, control of a capability or behavior of an agent, a task execution strategy, a scope of external interactions, or a fallback option for a task failure.
[0088] The policy for an agent or the IA policy information may be represented by the attribute “PolicyForAgent” in Table 2. Example attributes of the policy for the agent is shown in Table 3. Table 3. Example attributes of a policy for an agent
[0089] For example, the policy for the agent may comprise a managed activation scope indicating a scope of capabilities that may be activated or deactivated based on such as on / off control, condition-based activation / deactivation in terms of location, time or a specific type of task. The managed capabilities may comprise the IA capabilities as discussed above such as supported consumer services (e.g., analytics output) from other Network Functions or other service from same Network Function, the capability to use external tools / service (API) , the capability to interact with other agents, the capability to use multiple different back-end LLM or LLM based models. The managed activation scope may be represented by the attribute “managedActivationScope” as shown in Table 3.
[0090] The attribute “executionStrategy” indicates priority of task execution. It may be a list of ENUM bestPerformance, balanced, bestEnergySaving (hence less interaction with another agent) . The attribute “scopeOfExternalInteraction” indicates a filter about the scope per GeoLocation, per Network Function list, or AgentType, etc. The attribute “fallbackCapabilities” indicates when a sub task request failed, the fallback option for addressing the sub-task.
[0091] In some embodiments, the policy for an agent, i.e., the attribute “PolicyForAgent” may be part of the IA provisioning information.
[0092] In some embodiments, the first apparatus 110 may receive the IA policy information from a third apparatus that transmits the request to the first apparatus or a mobile network operator (MNO) . In some embodiments, the IA provisioning information may further comprise IA activation or deactivation configuration information. The IA activation or deactivation configuration information may comprise activation or deactivation of at least one of: an agent, a capability of an agent, a network function (NF) service, or a model of an agent.
[0093] In an example, a data type “ManagedActivation” may represent a capability to activate / deactivate Intelligent Agent’s capability, managed activation / deactivation scope for the agent capability, e.g., on / off control, condition based activate / deactivate in terms of location, time, specific type of task, capabilities like. The data type “ManagedActivation” may include example attributes of the data type “ManagedActivation” is shown in Table 4. Table 4. Example attributes of the data type “ManagedActivation”
[0094] The attribute “activationNFservice” indicates the Agent may consume services from other Network Functions, or other service from same Network Function. It may be a list of Service, or “ALL” indicates all NF services.
[0095] The attribute “activationAsPerLocation” indicates a location constraint that the Agent may be activated only when the task fulfillment relates to a specific geographical area. It may be a list of Location types (e.g., polygons, centers and radiuses, CellIDs) .
[0096] The attribute “activationAsPerNF” indicates an NF on which the agent may be activated. It may be a list of NF instances of NF types. The attribute “activationAsPerTime” indicates a time constraint that the agent may be activated only during the indicated time frame. It may be a list of time period. The attribute “activatedToolService” indicates using external tools or services. It may be a list of Tools, Service activated for the Agent.
[0097] The attribute “maxNumberOfExternalAgent” indicates maximum number of external Intelligent Agents for a given Request. It may be an Integer. Zero indicates interacting with external Agent is deactivated.
[0098] The attribute “activatedModel” indicates using multiple different back-end LLM or LLM based model. It may be a list of LLM or LLM based AIMLModel activated to support the Agent. The attribute “deactiveatedModel” indicates of blocking use of different back-end LLM or LLM based model. It may be a list of LLM or LLM based AIMLModel blocked to support the Agent.
[0099] In some embodiments, the first apparatus 110 may determine an agentic job corresponding to the execution plan. For example, the agentic job may represent the request. The first apparatus 110 may maintain agentic job information of the agentic job. The agentic job information may be used to manage the execution plan.
[0100] In some embodiments, the agentic job information may comprise at least one of: a job identification, a job status, an execution progress, a job request, a job context, or job report control. In some embodiments, at least a part of the job identification is configured as common identification of the at least one IAs.
[0101] In an example, as shown in FIG. 4, the information model 400 may include an IOC “AgenticJob” 405 for the agentic job information. Table 5 shows example attributes in agentic job information, for example in the IOC “AgenticJob” 405. Table 5. Example attributes of the agentic job information
[0102] The attributes “administrativeState” and “operationalState” are to maintain the status of the job. The attribute “jobId” may be used to associate tasks from multiple “IntelligentAgent” instances. The “jobId” may be included when reporting job execution status to allow a MnS consumer to associate received status report for the same request.
[0103] The attribute “executionProgress” indicates the current status and control information element, which is to start, hold-on, resume, cancel the job. The attribute “jobRequest” indicates the request from consumer. It may be the task or sub-tasks from the consumer.
[0104] The attribute “jobContext” indicates the job context during the job execution. It may indicate it is a sub-task as part of another task, or memory information to job, etc. The attribute “jobReportCtrl” indicates how report of job is generated. It may indicate report level job execution details, e.g., verbose, brief (default) , general, etc.
[0105] In some example embodiments, the first apparatus 110 may transmit a response for the request to a third apparatus that transmits the request to the first apparatus 110. The response may include an agentic job report. The agentic job report may be used to indicate the execution of the request to the third apparatus.
[0106] In some embodiments, the agentic job report may include at least one of: a job identification, a job status, a job result, or a job failure cause. As an example, the information model 400 may include an IOC “AgentJobReport” 402 representing the agentic job repot. Table 6 shows example attributes of the agentic job report. Table 6. Example attributes of the agentic job report
[0107] The attribute “jobId” may be used to associate report with the task. The attribute “jobStatus” indicate the final status of the job execution, it may be successful, failed, partially successful. The attribute “jobResults” indicates the detailed results of the report as per the configuration of the report. It may include different level of details. The attribute “jobFailureCause” indicates possible failure cause when the status is failed or partially successful. The cause maybe be an enumeration: 1) . job execution too late, 2) . insufficient data and so on.
[0108] In some embodiments, after receiving a fulfilment report from each of the at least one IA, the first apparatus 110 may store task planning or execution metadata corresponding to the request and an identifier of the at least one IA. For example, once the last IA provides its task fulfillment report, IAOF stores task metadata, task planning metadata and identifiers of the IAs for future usage.
[0109] As discussed above, in some example embodiments, the first apparatus 110 may be implemented in the AI / ML inference function. To this end, a new attribute “capabilityActivation” may be added to IOC “AIMLInferenceFunction” 401 as shown in FIG. 4. Table 7 shows example attributes in the IOC “AIMLInferenceFunction” 401. The new attribute “capabilityActivation” indicates the agent capability is enabled or not. It may be a list of Boolean when there is a list of Agents supported by the Inference Function. Table 7. Example attributes of the agentic job report
[0110] Some example embodiments are described above. An example general procedure is now described. In an example, the IAOF receives a consumer / operator prompt and, using a model (e.g., LLM / LMM / SLM) transforms it into an actionable network task. If the task can be mapped to CRUD operations, the IAOF may select and apply these CRUD operations. If the task cannot be mapped to CRUD operations, the IAOF decomposes the task into sub-tasks. The IAOF discovers and selects IAs for sub-task fulfillment and plans task fulfillment, including, but not limited to an order of sub-task fulfillment, deadlines for sub-task fulfillment, etc. IAOF shares sub-task fulfillment instructions to each network entity hosting the selected IA (s) of the agentic framework. The IAs selected for sub-task fulfillment act per task fulfillment plan and report upon task accomplishment to IAOF. Once the last IA provides its task fulfillment report, IAOF stores the task metadata, task planning metadata and identifiers of IAs for future usage.
[0111] To better understand the solution for IA orchestration, some example procedures are now described.
[0112] FIG. 5 illustrates a signaling flow of an example procedure 500 between a consumer and a producer concerning the lifecycle of a task or intent in accordance with some example embodiments of the present disclosure.
[0113] The consumer 501 of MnS for the AI / ML inference (also referred to as the Inference MnS consumer 501) initiates 505 a network request for a generative inference. It may be for a specific network task or an intent for a network request. The producer 502 of MnS for the AI / ML inference (also referred to as the Inference MnS Producer 502) checks 508 if an agent is needed for example by evaluating the raw input / prompt from the consumer. For example, if the task or the intent can be done via mapping to a one direct CRUD operation or a set of CRUD operations without interacting with related network entities or services, there is no need to initiate any intelligent agent. If an agent is not needed, the Inference MnS Producer 502 may indicate the Inference MnS consumer 501 to generate results directly, for example with external tools.
[0114] If the evaluation result is to have at least one Intelligent Agent, the Inference MnS Producer 502 instantiates the Intelligent Agent Instance. For example, the Inference MnS Producer 502 creates 515 an intelligent agent Managed Object Instance (MOI) . Alternatively, the Intelligent Agent instance may be installed by the system.
[0115] The Inference MnS Producer 502 (with help from LLM or LLM based model) decomposes 520 the request and breaks down the request to an execution plan. The execution plan may include details such as whether external Intelligent Agent needed or not, what kind of Intelligent Agent is needed, how many Intelligent Agents are needed, what service (s) to be consumed, what tools to be used and how they can be accessed.
[0116] The Inference MnS Producer 502 sends 525 back an initial response to the Inference MnS Consumer 501 to indicate that the request is accepted and may further indicate if more time is needed. Optionally, the Inference MnS Producer 502 may indicate (e.g., in natural language) the sub-tasks relating to the received prompt.
[0117] The Inference MnS Producer 502 may use 530 existing Intelligent Agent instances from other producers. Alternatively, the Inference MnS Producer 502 may request to create the Intelligent Agent Instance directly for the sub-task. Alternatively, the Inference MnS Producer 502 may send a normal generative request for the sub-task. It is up to the related MnS Producer to decide an agent is needed or not and create an Intelligent Agent when needed and negotiate with the primary agent (which is also referred to as a planning agent) .
[0118] The Inference MnS Producer 502 negotiates 535 with other Intelligent Agent service producer for subtask assignment. The Inference MnS Producer 502 orchestrates 540 the interactions between the primary agent and assigned agents to execute the assigned subtasks by interacting 545 with the assigned agents for sub-task execution.
[0119] The Inference MnS Consumer 501 may receive 550 some notifications during the job execution. The Inference MnS Producer 502 may send 555 a job report to the Inference MnS Consumer 501 when job execution is completed or failed.
[0120] FIG. 6 illustrates a signaling flow of an example procedure 600 for agent discovery and configuration in accordance with some example embodiments of the present disclosure.
[0121] At a stage 601 of post deployment commissioning, the Inference MnS Consumer 501 provisions 605 general capabilities (for example the AgentCapability mentioned above) to the Inference MnS Producer 502. The Inference MnS Consumer 501 provisions 610 general activation / deactivation configuration (for example the ManagedActivation mentioned above) to the Inference MnS Producer 502. The Inference MnS Consumer 501 provisions 615 capabilities related to use of tools to the Inference MnS Producer 502. The capabilities related to use of tools may include for example a tool type, a tool level, a tool policy. The Inference MnS Consumer 501 provisions 620 a policy for an agent to the Inference MnS Producer 502. The Inference MnS Consumer 501 provisions 625 a collaboration policy of using other agents to the Inference MnS Producer 502. The collaboration policy may include for example a policy for an agent, a context and conditions. The Inference MnS Consumer 501 provisions 630 LLM or LLM based model to the Inference MnS Producer 502.
[0122] At a stage 602 of run time provisioning, the Inference MnS Consumer 501 updates 635 a decomposition policy (for example the AgentCapability) to the Inference MnS Producer 502. The Inference MnS Consumer 501 updates 640 tools (for example a tool type, a tool level, a tool policy) to the Inference MnS Producer 502. The Inference MnS Consumer 501 updates 645 general activation / deactivation configuration (for example the ManagedActivation mentioned above) to the Inference MnS Producer 502. The Inference MnS Consumer 501 updates 650 a collaboration policy of using other agents to the Inference MnS Producer 502. The collaboration policy may include for example a policy for an agent, a context and conditions. The Inference MnS Consumer 501 updates 655 LLM or LLM based model to the Inference MnS Producer 502.
[0123] Now some use cases are described. In telecommunication systems, for a simple service request of tasks, which may be mapped to direct CRUD operation or set of CRUD Operations and no two-way interaction needed, there is no need for an agent. For a service request or tasks more complex than these, the LLM (or LLM based) agent may be needed. The agent may decompose the request or tasks to an action plan or an execution plan. Each item may be a simpler task which can be mapped to a direct CRUD operation or a set of CRUD operations, and other items may require interacting with / consuming services from other network function (s) , using external tools, or interacting with agent (s) in other network function (s) .
[0124] In the following two example use cases are described for the purpose of illustrating the application of agentic frameworks in the telecommunication domain. The described use cases benefit from the solution proposed in the present disclosure in terms of managing different agents, their capabilities and interactions.
[0125] FIG. 7 illustrates an example diagram 700 of a use case example where intelligent agents are aiding network management and optimization in accordance with some example embodiments of the present disclosure. In this example, agents aid network management and optimization, which is intent based management.
[0126] In this scenario, a human operator 701 may express an intent e.g. “I wish to get the RAN energy consumption in area X below Y kwh” . Such a natural language intent is given to the planning intelligent agent 702. The planning intelligent agent 702 leverages an LLM to comprehend the given human intent. It derives that different energy saving tasks need to be performed across certain area of interest. Based on the registered capabilities or Intelligent Agent performing specific task (e.g. power level optimization, RAN and Core network energy consumption saving) , the planning intelligent agent 702 derives which intelligent agents are the most suitable and shall be involved in intent fulfillment.
[0127] In this example, the planning intelligent agent 702 may select three intelligent agents, each responsible for a specific task. For example, the intelligent agent 703 is responsible for transmit power adjustment, the intelligent agent 704 is responsible for cell switch on / off and the intelligent agent 703 is responsible for selection of most energy efficient UPF and according to traffic routing. The planning intelligent agent 702 may monitor the execution of other intelligent agents. Along with further inputs and interactions with the environment, the planning intelligent agent 702 may determine if the intent of the operator 701 has been fulfilled. The feedback may be sent to the operator informing about the intent fulfillment or failure.
[0128] FIG. 8 illustrates an example diagram of a use case example where intelligent agents are tailored to consumer-oriented tasks in accordance with some example embodiments of the present disclosure. FIG. 8 shows an example of consumer-based assistance in daily activities. In this example, agents may perform specific tasks aiding the human in completion of his / her activities (e.g. fitness coach, cooking teacher, sightseeing guidance etc. ) .
[0129] In such a use case, a human user 801 may express tasks by using a UE 802. The tasks may include e.g. creating vacation itinerary in specific location and asking assistance in all related facilities selection and booking, e.g. travel tickets, accommodation, restaurants, visiting landmarks, etc. To complete such complex tasks, the planning intelligent agent 803 may determine that different agents may be needed, each specialized for a part of the complete task. For example, an intelligent agent may be specialized for accommodation and restaurant selection based on information on availability, rating, customer’s budget, target quality level, etc. Another intelligent agent may be specialized for selecting the most appropriate landmarks and attractions to visit, based on customer preferences, age, interests, budget, etc. All such intelligent agents need to collaborate between each other such that the complex task of vacation planning and execution is fulfilled.
[0130] The planning intelligent agent 803 selects the most suitable intelligent agents 804, 805 and 806, given the consumer’s task description as well as capabilities of available agents including their interaction capabilities. For example, in this case, only intelligent agents capable of assisting in desired vacation location are suitable, also interacting with each other such that booking times and locations of restaurant are in sync with booking times of attractions to visit. The planning intelligent agent 803 may provide the feedback to the user if the required task can be fulfilled given the constraints and availability of intelligent agents.
[0131] To sum up, in the present disclosure, management means are proposed to enable control of decomposition of a complex task into smaller tasks and their execution towards complex task fulfillment, i.e. control of planning performed by Intelligent Agent. This comprises providing planning policies for control of planning process. The proposed management means may include exposure and registration of description of a complex task for which the intelligent agent is capable to decompose into simpler tasks.
[0132] The proposed management means may include exposure and registration of task decomposition metadata. The exposure and registration of task decomposition metadata may include examples of previous complex task decompositions, involved smaller tasks, task feasibility indicator (i.e., based on CRUD operations and / or intelligent agents) , task performance level indicators when applying the plan derived by the intelligent agent (s) , cost of different task fulfillment plans in terms of monetary cost, energy spending, interface signaling etc.
[0133] The proposed management means may include defining planning policies that steer and control the planning process towards task accomplishment. For example, the following aspects may be defined: which intelligent agents with which capabilities is to be included or excluded from the plan, e.g., filtered by type, geo location, etc, maximum number of agents to be involved in the plan, which LLM / LMM / SLM models is to be part of intelligent agents to be included in the plan, priority of tasks within the plan, how the mapping between the task, agent and model should look like for example one-to-one, one-to-many, etc.
[0134] In the telecommunication system, intelligent agents from different vendors may not only perform different tasks or nuances of tasks, but may also be based on completely different models having different capabilities and requirements related to needed inputs and providing different outputs. The present disclosure proposes a mechanism for enabling management of an agentic framework in the communication network. In this way, intelligent agents from different vendors cab collaborate to perform a task.
[0135] FIG. 9 shows a flowchart of an example method 900 implemented at a first apparatus in accordance with some example embodiments of the present disclosure. For the purpose of discussion, the method 900 will be described from the perspective of the first apparatus 110 in FIG. 1.
[0136] At block 910, in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, the first apparatus 110 determines based on the request, an execution plan to be executed by at least one IA.
[0137] At block 920, the first apparatus 110 transmits an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.
[0138] In some example embodiments, the method 900 further comprises: determining the at least one IA by using a first IA hosted in the first apparatus, or determining, based on the request, the at least one IA comprising the first IA hosted in the first apparatus.
[0139] In some example embodiments, determining the at least one IA comprises: discovering a plurality of IAs and IA capability information of the plurality of IAs; and selecting, based on the request and the IA capability information, the at least one IA from the plurality of IAs.
[0140] In some example embodiments, the method 900 further comprises: receiving IA provisioning information, and wherein determining the execution plan to be executed by the at least one IA is further based on the IA provisioning information.
[0141] In some example embodiments, the IA provisioning information comprises IA capability information comprising at least one of: information of at least one task supported by an agent, a capability of a base model for an agent, a capability to use external knowledge or tool, a capability to collaborate with a further agent, a context length of an IA, or a capability for verifying and refining an output from a further agent.
[0142] In some example embodiments, receiving the IA provisioning information comprises: receiving the IA capability information from at least one of a third apparatus that transmits the request to the first apparatus or a fourth apparatus responsible for an exposure or discovery service of the IA capability information.
[0143] In some example embodiments, the IA provisioning information comprises IA agentic information comprising at least one of: an agent capability, a policy for an agent, an agent type, an agent accountability score, or a potential collaboration agent or a list of potential collaboration agents.
[0144] In some example embodiments, the IA provisioning information comprises IA policy information comprising at least one of: a managed activation / deactivation scope for agent capabilities, control of a capability or behavior of an agent, a task execution strategy, a scope of external interactions, or a fallback option for a task failure.
[0145] In some example embodiments, receiving the IA provisioning information comprises: receiving the IA policy information from a mobile network operator (MNO) or a third apparatus that transmits the request to the first apparatus.
[0146] In some example embodiments, the IA provisioning information comprises IA activation or deactivation configuration information comprising activation or deactivation of at least one of: an agent, a capability of an agent, a network function (NF) service, or a model of an agent.
[0147] In some example embodiments, the method 900 further comprises: receiving an update of the IA provisioning information during runtime of the execution plan.
[0148] In some example embodiments, the method 900 further comprises: determining an agentic job corresponding to the execution plan; and maintaining agentic job information of the agentic job.
[0149] In some example embodiments, the agentic job information comprises at least one of:job identification, job status, execution progress, a job request, job context, or job report control.
[0150] In some example embodiments, at least a part of the job identification is configured as a common identification of the at least one IA.
[0151] In some example embodiments, the method 900 further comprises: transmitting a response for the request to a third apparatus that transmits the request to the first apparatus, the response comprising an agentic job report.
[0152] In some example embodiments, the agentic job report comprises at least one of: job identification, job status, a job result, or a job failure cause.
[0153] In some example embodiments, the execution task is one of a plurality of execution tasks of the execution plan, and determining, based on the request, the execution plan to be executed by the at least one IA comprises: decomposing the request into the plurality of execution tasks having at least one of an execution dependency or execution constraint.
[0154] In some example embodiments, the method 900 further comprises: determining, based on the request, whether an IA is needed for responding to the request.
[0155] In some example embodiments, determining, based on the request, whether the IA is needed for responding to the request comprises: determining whether the request can map to at least one of: a set of create, read, update, delete (CRUD) operations, a direct NF service or external tool calling.
[0156] In some example embodiments, the method 900 further comprises: after receiving a fulfilment report from each of the at least one IA, storing task planning or execution metadata corresponding to the request and an identifier of the at least one IA.
[0157] In some example embodiments, the method 900 further comprises: transmitting the task planning or execution metadata to a further apparatus responsible for an exposure or discovery service of the task planning or execution metadata.
[0158] In some example embodiments, the at least one IA is based on at least one of: a large language model (LLM) , a large multimodal model (LMM) , or a small language model (SLM) .
[0159] In some example embodiments, first apparatus is or comprised 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.
[0160] In some example embodiments, a first apparatus capable of performing any of the method 900 (for example, the first apparatus 110 in FIG. 1) may comprise means for performing the respective operations of the method 900. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module. The first apparatus may be implemented as or included in the first apparatus 110 in FIG. 1.
[0161] In some example embodiments, the first apparatus comprises means for in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determining based on the request, an execution plan to be executed by at least one IA; and means for transmitting an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.
[0162] In some example embodiments, the first apparatus further comprises: means for determining the at least one IA by using a first IA hosted in the first apparatus, or means for determining, based on the request, the at least one IA comprising the first IA hosted in the first apparatus.
[0163] In some example embodiments, the means for determining the at least one IA comprises: means for discovering a plurality of IAs and IA capability information of the plurality of IAs; and means for selecting, based on the request and the IA capability information, the at least one IA from the plurality of IAs.
[0164] In some example embodiments, the first apparatus further comprises: means for receiving IA provisioning information, and wherein determining the execution plan to be executed by the at least one IA is further based on the IA provisioning information.
[0165] In some example embodiments, the IA provisioning information comprises IA capability information comprising at least one of: information of at least one task supported by an agent, a capability of a base model for an agent, a capability to use external knowledge or tool, a capability to collaborate with a further agent, a context length of an IA, or a capability for verifying and refining an output from a further agent.
[0166] In some example embodiments, the means for receiving the IA provisioning information comprises: means for receiving the IA capability information from at least one of a third apparatus that transmits the request to the first apparatus or a fourth apparatus responsible for an exposure or discovery service of the IA capability information.
[0167] In some example embodiments, the IA provisioning information comprises IA agentic information comprising at least one of: an agent capability, a policy for an agent, an agent type, an agent accountability score, or a potential collaboration agent or a list of potential collaboration agents.
[0168] In some example embodiments, the IA provisioning information comprises IA policy information comprising at least one of: a managed activation / deactivation scope for agent capabilities, control of a capability or behavior of an agent, a task execution strategy, a scope of external interactions, or a fallback option for a task failure.
[0169] In some example embodiments, means for receiving the IA provisioning information comprises: means for receiving the IA policy information from a mobile network operator (MNO) or a third apparatus that transmits the request to the first apparatus.
[0170] In some example embodiments, the IA provisioning information comprises IA activation or deactivation configuration information comprising activation or deactivation of at least one of: an agent, a capability of an agent, a network function (NF) service, or a model of an agent.
[0171] In some example embodiments, the first apparatus further comprises: means for receiving an update of the IA provisioning information during runtime of the execution plan.
[0172] In some example embodiments, the first apparatus further comprises: means for determining an agentic job corresponding to the execution plan; and means for maintaining agentic job information of the agentic job.
[0173] In some example embodiments, the agentic job information comprises at least one of:job identification, job status, execution progress, a job request, job context, or job report control.
[0174] In some example embodiments, at least a part of the job identification is configured as a common identification of the at least one IA.
[0175] In some example embodiments, the first apparatus further comprises: means for transmitting a response for the request to a third apparatus that transmits the request to the first apparatus, the response comprising an agentic job report.
[0176] In some example embodiments, the agentic job report comprises at least one of: job identification, job status, a job result, or a job failure cause.
[0177] In some example embodiments, the execution task is one of a plurality of execution tasks of the execution plan, and the means for determining, based on the request, the execution plan to be executed by the at least one IA comprises: means for decomposing the request into the plurality of execution tasks having at least one of an execution dependency or execution constraint.
[0178] In some example embodiments, the first apparatus further comprises: means for determining, based on the request, whether an IA is needed for responding to the request.
[0179] In some example embodiments, the means for determining, based on the request, whether the IA is needed for responding to the request comprises: means for determining whether the request can map to at least one of: a set of create, read, update, delete (CRUD) operations, a direct NF service or external tool calling.
[0180] In some example embodiments, the first apparatus further comprises: means for after receiving a fulfilment report from each of the at least one IA, storing task planning or execution metadata corresponding to the request and an identifier of the at least one IA.
[0181] In some example embodiments, the first apparatus further comprises: means for transmitting the task planning or execution metadata to a further apparatus responsible for an exposure or discovery service of the task planning or execution metadata.
[0182] In some example embodiments, the at least one IA is based on at least one of: a large language model (LLM) , a large multimodal model (LMM) , or a small language model (SLM) .
[0183] In some example embodiments, first apparatus is or comprised 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.
[0184] FIG. 10 is a simplified block diagram of a device 1000 that is suitable for implementing example embodiments of the present disclosure. The device 1000 may be provided to implement a communication device, for example, the apparatus in FIG. 1. As shown, the device 1000 includes one or more processors 1010, one or more memories 1020 coupled to the processor 1010, and one or more communication modules 1040 coupled to the processor 1010.
[0185] The communication module 1040 is for bidirectional communications. The communication module 1040 has one or more communication interfaces to facilitate communication with one or more other modules or devices. The communication interfaces may represent any interface that is necessary for communication with other network elements. In some example embodiments, the communication module 1040 may include at least one antenna.
[0186] The processor 1010 may be of any type suitable to the local technical network and may include one or more of the following: general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and processors based on multicore processor architecture, as non-limiting examples. The device 1000 may have multiple processors, such as an application specific integrated circuit chip that is slaved in time to a clock which synchronizes the main processor.
[0187] The memory 1020 may include one or more non-volatile memories and one or more volatile memories. Examples of the non-volatile memories include, but are not limited to, a Read Only Memory (ROM) 1024, an electrically programmable read only memory (EPROM) , a flash memory, a hard disk, a compact disc (CD) , a digital video disk (DVD) , an optical disk, a laser disk, and other magnetic storage and / or optical storage. Examples of the volatile memories include, but are not limited to, a random-access memory (RAM) 1022 and other volatile memories that will not last in the power-down duration.
[0188] A computer program 1030 includes computer executable instructions that are executed by the associated processor 1010. The instructions of the program 1030 may include instructions for performing operations / acts of some example embodiments of the present disclosure. The program 1030 may be stored in the memory, e.g., the ROM 1024. The processor 1010 may perform any suitable actions and processing by loading the program 1030 into the RAM 1022.
[0189] The example embodiments of the present disclosure may be implemented by means of the program 1030 so that the device 1000 may perform any process of the disclosure as discussed with reference to FIG. 2 to FIG. 9. The example embodiments of the present disclosure may also be implemented by hardware or by a combination of software and hardware.
[0190] In some example embodiments, the program 1030 may be tangibly contained in a computer readable medium which may be included in the device 1000 (such as in the memory 1020) or other storage devices that are accessible by the device 1000. The device 1000 may load the program 1030 from the computer readable medium to the RAM 1022 for execution. In some example embodiments, the computer readable medium may include any types of non-transitory storage medium, such as ROM, EPROM, a flash memory, a hard disk, CD, DVD, and the like. The term “non-transitory, ” as used herein, is a limitation of the medium itself (i.e., tangible, not a signal) as opposed to a limitation on data storage persistency (e.g., RAM vs. ROM) .
[0191] FIG. 11 shows an example of the computer readable medium 1100 which may be in form of CD, DVD or other optical storage disk. The computer readable medium 1100 has the program 1030 stored thereon.
[0192] Generally, various embodiments of the present disclosure may be implemented in hardware or special purpose circuits, software, logic or any combination thereof. Some aspects may be implemented in hardware, and other aspects may be implemented in firmware or software which may be executed by a controller, microprocessor or other computing device. Although various aspects of embodiments of the present disclosure are illustrated and described as block diagrams, flowcharts, or using some other pictorial representations, it is to be understood that the block, apparatus, system, technique or method described herein may be implemented in, as non-limiting examples, hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.
[0193] Some example embodiments of the present 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, such as those included in program modules, being executed in a device on a target physical or virtual processor, to carry out any of the methods as described above. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, or the like that perform particular tasks or implement particular abstract data types. The functionality of the program modules may be combined or split between program modules as desired in various embodiments. Machine-executable instructions for program modules may be executed within a local or distributed device. In a distributed device, program modules may be located in both local and remote storage media.
[0194] Program code for carrying out methods of the present 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 the program code, when executed by the processor or controller, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may execute entirely on a machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
[0195] In the context of the present disclosure, the computer program code or related data may be carried by any suitable carrier to enable the device, apparatus or processor to perform various processes and operations as described above. Examples of the carrier include a signal, computer readable medium, and the like.
[0196] The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable medium may include but not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the computer readable storage medium would include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random-access memory (RAM) , a read-only memory (ROM) , an erasable programmable read-only memory (EPROM or Flash memory) , an optical fiber, a portable compact disc read-only memory (CD-ROM) , an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0197] Further, although operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, although several specific implementation details are contained in the above discussions, these should not be construed as limitations on the scope of the present disclosure, but rather as descriptions of features that may be specific to particular embodiments. Unless explicitly stated, certain features that are described in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, unless explicitly stated, various features that are described in the context of a single embodiment may also be implemented in a plurality of embodiments separately or in any suitable sub-combination.
[0198] Although the present disclosure has been described in languages specific to structural features and / or methodological acts, it is to be understood that the present disclosure defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.
Claims
1.A first apparatus comprising:at least one processor; andat least one memory storing instructions that, when executed by the at least one processor, cause the first apparatus to:in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determine, based on the request, an execution plan to be executed by at least one IA; andtransmit an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.2.The first apparatus of claim 1, wherein the first apparatus is further caused to:determine the at least one IA by using a first IA hosted in the first apparatus, ordetermine, based on the request, the at least one IA comprising the first IA hosted in the first apparatus.3.The first apparatus of claim 2, wherein determining the at least one IA comprises:discovering a plurality of IAs and IA capability information of the plurality of IAs; andselecting, based on the request and the IA capability information, the at least one IA from the plurality of IAs.4.The first apparatus of any of claims 1 to 3, wherein the first apparatus is further caused to:receive IA provisioning information, andwherein determining the execution plan to be executed by the at least one IA is further based on the IA provisioning information.5.The first apparatus of claim 4, wherein the IA provisioning information comprises IA capability information comprising at least one of:information of at least one task supported by an agent,a capability of a base model for an agent,a capability to use external knowledge or tool,a capability to collaborate with a further agent,a context length of an IA, ora capability for verifying and refining an output from a further agent.6.The first apparatus of claim 4, wherein receiving the IA provisioning information comprises:receiving the IA capability information from at least one of a third apparatus that transmits the request to the first apparatus or a fourth apparatus responsible for an exposure or discovery service of the IA capability information.7.The first apparatus of any of claims 4 to 6, wherein the IA provisioning information comprises IA agentic information comprising at least one of:an agent capability,a policy for an agent,an agent type,an agent accountability score, ora potential collaboration agent or a list of potential collaboration agents.8.The first apparatus of any of claims 4 to 7, wherein the IA provisioning information comprises IA policy information comprising at least one of:a managed activation / deactivation scope for agent capabilities,control of a capability or behavior of an agent,a task execution strategy,a scope of external interactions, ora fallback option for a task failure.9.The first apparatus of claim 8, wherein receiving the IA provisioning information comprises:receiving the IA policy information from a mobile network operator (MNO) or a third apparatus that transmits the request to the first apparatus.10.The first apparatus of any of claims 4 to 9, wherein the IA provisioning information comprises IA activation or deactivation configuration information comprising activation or deactivation of at least one of:an agent,a capability of an agent,a network function (NF) service, ora model of an agent.11.The first apparatus of any of claims 4 to 10, wherein the first apparatus is further caused to:receive an update of the IA provisioning information during runtime of the execution plan.12.The first apparatus of any of claims 1 to 11, wherein the first apparatus is further caused to:determine an agentic job corresponding to the execution plan; andmaintain agentic job information of the agentic job.13.The first apparatus of claim 12, wherein the agentic job information comprises at least one of:job identification, job status, execution progress, a job request, job context, or job report control.14.The first apparatus of claim 13, wherein at least a part of the job identification is configured as a common identification of the at least one IA.15.The first apparatus of any of claims 4 to 14, wherein the first apparatus is further caused to:transmit a response for the request to a third apparatus that transmits the request to the first apparatus, the response comprising an agentic job report.16.The first apparatus of claim 15, wherein the agentic job report comprises at least one of:job identification, job status, a job result, or a job failure cause.17.The first apparatus of any of claims 1 to 16, wherein the execution task is one of a plurality of execution tasks of the execution plan, and determining, based on the request, the execution plan to be executed by the at least one IA comprises:decomposing the request into the plurality of execution tasks having at least one of an execution dependency or execution constraint.18.The first apparatus of any of claims 1 to 17, wherein the first apparatus is further caused to:determine, based on the request, whether an IA is needed for responding to the request.19.The first apparatus of claim 18, wherein determining, based on the request, whether the IA is needed for responding to the request comprises:determining whether the request can map to at least one of: a set of create, read, update, delete (CRUD) operations, a direct NF service or external tool calling.20.The first apparatus of any of claims 1 to 19, wherein the first apparatus is further caused to:after receiving a fulfilment report from each of the at least one IA, store task planning or execution metadata corresponding to the request and an identifier of the at least one IA.21.The first apparatus of claim 20, wherein the first apparatus is further caused to:transmit the task planning or execution metadata to a further apparatus responsible for an exposure or discovery service of the task planning or execution metadata.22.The first apparatus of any of claims 1 to 21, wherein the at least one IA is based on at least one of:a large language model (LLM) ,a large multimodal model (LMM) , ora small language model (SLM) .23.The first apparatus of any of claims 1 to 22, wherein first apparatus is or comprised 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:in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determining based on the request, an execution plan to be executed by at least one IA; andtransmitting an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.25.A first apparatus comprising:means for in accordance with a determination that an intelligent agent (IA) is needed for responding to a request, determining based on the request, an execution plan to be executed by at least one IA; andmeans for transmitting an execution task of the execution plan to a second apparatus hosting a target IA of the at least one IA.26.A computer readable medium comprising instructions stored thereon for causing an apparatus at least to perform the method of claim 24.