Referee and bookkeeping for interworking between smart agents
A referee service verifies and enforces rules for AI agents, preventing cheating and ensuring fair interworking by validating their actions, thus maintaining reliable collaboration.
Patent Information
- Application Number
- PCT/CN2025/094045
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2026-11-12
AI Technical Summary
AI agents in communication systems may cheat each other during interworking, leading to unfair advantages and disrupting normal collaboration, particularly in scenarios like self-driving cars negotiating right-of-way.
Implement a referee service that verifies actions of AI agents participating in a task, ensuring compliance with predefined rules and preventing cheating by using a referee entity that enforces interworking rules and bookskeeping the process.
Prevents cheating among AI agents, ensuring fair and reliable interworking by validating actions against predefined rules, thereby maintaining expected collaboration outcomes.
Smart Images

Figure CN2025094045_12112026_PF_FP_ABST
Abstract
Description
REFEREE AND BOOKKEEPING FOR INTERWORKING BETWEEN SMART AGENTSFIELD
[0001] Various example embodiments of the present disclosure generally relate to the field of telecommunications and in particular, to a referee service for a telecommunications network.BACKGROUND
[0002] An artificial intelligence (AI) agent is an automated entity that senses and responds to its environment and takes actions to achieve its goal. With the development of AI technology, AI agents play a greater role in more fields of communication systems and promote the intelligent development of the communication industry. Collaboration and communication among AI agents become a topic in the third-generation partnership project (3GPP) standards. AI agents may be interworking to achieve common objectives or compete against each other so as to optimize their own achievements. One example use case may be collaboration among self-driving cars. For example, two self-driving cars may approach an intersection and negotiate right-of-way using vehicle-to-vehicle (V2V) communication. However, during interworking of AI agents, these AI agents may cheat each other to take advantage over each other.SUMMARY
[0003] According to some aspects, there is provided the subject matter of the independent claims. Some further aspects are defined in the dependent claims.
[0004] In a first aspect, there is provided a device comprising means for obtaining, from an AI agent on a user equipment, a first action generated by the AI agent, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least the AI agent; means for sending, to a referee service of a communication network, a request for verification of the first action, the request comprising information related to the first action, wherein the request causes the referee service to verify the first action; and means for receiving a response to the request, wherein the response indicates a result of the verification.
[0005] In a second aspect, there is provided a device, method, computer-readable medium, computer program product for receiving, from a user equipment, a request for verification of a first action, the request comprising information related to the first action, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least an AI agent on the user equipment; in response to receiving the request, performing a referee service to verify the first action; and sending, to the user equipment, a response to the request, wherein the response indicates a result of the verification.
[0006] It is to be understood that the Summary 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
[0007] Some example embodiments will now be described with reference to the accompanying drawings, where:
[0008] FIG. 1 illustrates an example telecommunications network that includes a referee service for referring AI agents on UEs participating in a task according to some example embodiments;
[0009] FIG. 2 illustrates a signaling flow of an example process of a referee service in accordance with some example embodiments;
[0010] FIGS. 3 and 4 illustrate signaling flows of other processes of referee services in accordance with some example embodiments;
[0011] FIG. 5 illustrates a signaling flow of an example process of coordination between referees across networks in accordance with some example embodiments;
[0012] FIGS. 6 and 7 shows flowcharts of example methods of a referee service in accordance with some example embodiments; and
[0013] FIG. 8 illustrates a simplified block diagram of a device that is suitable for implementing example embodiments.
[0014] Throughout the drawings, the same or similar reference numerals represent the same or similar element.DETAILED DESCRIPTION
[0015] Principles 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.
[0016] 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.
[0017] 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.
[0018] 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.
[0019] 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.
[0020] As used herein, unless stated explicitly, performing a step “in response to A” or “responsive to A” does not indicate that the step is performed immediately after “A” occurs and one or more intervening steps may be included.
[0021] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting 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.
[0022] As used in this application, the term “circuitry” may refer to one or more or all of the following:
[0023] (a) hardware-only circuit implementations (such as implementations in analog, digital and / or quantum circuitry) and
[0024] (b) combinations of hardware circuit (s) and software, such as (as applicable) :
[0025] (i) a combination of analog, digital and / or quantum hardware circuit (s) with software / firmware and
[0026] (ii) any or all portions of hardware processor (s) (including digital and / or quantum processor (s) with software, and memory (ies) that work together to cause an apparatus, such as a mobile device, computing device, or server, to perform various functions) and
[0027] (c) any or all portions of hardware circuit (s) , such as a microprocessor (s) , processor (s) , and / or quantum processor (s) , that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation.
[0028] 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.
[0029] As used herein, the term “communication network” refers to a network following any suitable communication standards. Furthermore, the communication between a user device and a network device in the communication network may be performed according to any suitable generation communication protocols, wireless local network communication protocols, and / or any other protocols either currently known or to be developed in the future. Moreover, the communication may utilize any proper wireless communication technology currently known or to be developed in the future.
[0030] As used herein, the term “access node” (or AN) or “access network node” refers to a device in a communication network via which user equipment accesses a core network and receives services from the core network. The access network node can be a wireless access network node or a wireline access network node. Examples of the wireless access network nodes may include, but be not limited 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) , an Next Generation Radio Access Network Node (also referred to as an NG-RAN node) , 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, a RAN node (for example, a gNB) may include a Centralized Unit (CU) and one or more Distributed Units (DUs) connected to the CU. Examples of the wireless access network nodes may include, but be not limited to, a Wireline 5G Access Network (W-5GAN) node, for example, a Wireline 5G BBF Access Network (W-5GBAN) node or a Wireline 5G Cable Access Network (W-5GCAN) node.
[0031] As used herein, a core network device a core network may be any computing device or computing system that includes hardware (e.g., at least one processor and at least one memory) and software Examples of a core networks of a 4G telecommunication network may include, but be not limited to, an evolved Packet Data Gateway (ePGW) , a trusted wireless local area network (WLAN) access network (TWAN) node, a Home Subscriber Server (HSS) , Serving Gateway (SGW) , and Packet Data Network (PDN) Gateway (PGW) , , Examples of networks functions of core network devices may include, but be not limited to, an Access and Mobility Management Function (AMF) , a Session Management Function (SMF) , a Network Slice Selection Function (NSSF) , an Authentication Server Function (AUSF) , a Subscription Identifier De-concealing function (SIDF) , a Unified Data Management (UDM) , a Unified Data Repository (UDR) , a Security Edge Protection Proxy (SEPP) , a Network Exposure Function (NEF) , a Network Repository Function (NRF) , a User Plane Function (UPF) , a Location Management Function (LMF) , and / or a Policy Control Function (PCF) .
[0032] The term “user equipment” (or UE) refers to any device that is capable of wireless communication with a communication network. By way of example rather than limitation, user equipment may also be referred to as a communication device, terminal device, a Subscriber Station (SS) , a Portable Subscriber Station, a Mobile Station (MS) , or an Access Terminal (AT) . A user equipment may be 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. 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” , “terminal” , “user equipment” and “UE” may be used interchangeably.
[0033] The term “areferee agent” refers to an entity that serves as a referee for a given task and a set of agents. It can be considered as an instance of a referee function or the function itself.
[0034] In a communication network, AI agents may be deployed on the UEs and / or one or more NFs, to sense and respond to its environment and take actions to achieve its goal. The AI agent may need support of a large language model (LLM) or large multimodal model (LMM) , but the LLM or LMM may not be collocated with or to be a part of the AI agent. Such an AI agent will be referred to as a “LLM / LMM based agent” . An LLM / LMM based agent may be powered by LLM or LMM models and have advanced capabilities for reasoning, complex problem resolution, external tool usage, and / or the like.
[0035] The relationship between such an agent and an LLM or LMM (often at the backend) may be: (1) 1 to 1, or (2) many to many. In other words, from the same LLM or LMM, the capabilities and behaviors of different agents may be different, due to different configurations to the agents. The different agents may be for different types of tasks (or optimized for their types of tasks) , even though the intelligence is from the same LLM or LMM.
[0036] LLM / LMM based agents may be interworking to achieve common objectives or compete against each other so as to optimize their own achievements. During the interworking, these agents may cheat each other to take advantage over each other. Cheating behaviors may appear due to a stress added onto the agents or benefits (e.g., those rewards that may be acquired by a winning agent or lost by a losing agent) . It is needed to ensure that the AI agent in communication or collaboration is not cheating.
[0037] By way of example, LLM / LMM based agents or simply AI agents may be deployed and run locally in the UEs 110 and 120. The UEs 110 and 120 may be subscribers to the same network operator or different operators. A single UE 110 or 120 may have multiple agents, each dedicated to a specific task, and agents from different UEs 110 and 120 may need to collaborate on the same task. These agents may be owned by different human users. Cheating actions between the interacting agents among the UEs 110 and 120 may create issues for their human users. For example, an agent at a UE may break the rule and take a shortcut for the advantage of its human user when interworking with another agent at another UE. Such cheating may ruin the normal and expected interworking between the agents involved.
[0038] Example embodiments of the present disclosure relate to a referee service for a set of AI agents (or group of AI agents) hosted and running on UEs that are interworking on a task. In the referee service, a referee agent (shortly a referee) or a series of referees may be allocated for interworking between the set of AI agents. The referee service may be provided by a new NF or by an existing NF of a core network, a radio access network (RAN) , or an operation administration and maintenance (OAM) . An AI agent that is trusted by the interworking agents may act as a referee for the interworking agents, may enforce the interworking rule, and may bookkeep the interworking process. In this way, cheating actions of the interworking agents may be prevented, and the expected interworking may be achieved.
[0039] FIG. 1 shows an example telecommunication system 100 in which AI agents on UEs 110 and 120 such as smartphones or electric vehicle (e.g., hosted and running on the UEs 110 and 120) are interworking on a task under the supervision of a referee service 105 of a communication network 125 (hereinafter referred to as “network 125” ) according to some example embodiments. The AI agents on the UEs 110 and 120 (or UE1 and UE2) may be generative AI (GenAI) agents, LLM based agents or LMM based agents, which may be authenticated or authorized for accessing the network 125. The referee service 105 may be discovered by AI agents. The referee service may be selected and used for refereeing a set or group of AI agents interworking on (e.g., participating in) a specific task based on rules that the referee service needs to enforce. The rules may be related to the task and / or a use case the AI agents are interworking on (or participating in) . The referee service may be configured to track and react to the actions of the AI agents, for example, for auditing the actions and to ensure that the actions comply with regulatory requirements.
[0040] A referee service as described herein may be or may be implemented as a referee entity that includes a referee configured to verify actions generated by AI agents of interworking group (e.g., actions of AI agents that are working together or participating in a task) and the referee is provided as a service by the referee entity to other entities (e.g., other AI agents or UEs) . A referee entity may include a combination of hardware processing circuit (s) and software and / or firmware comprising machine-readable instructions, or software comprising machine-readable instructions that are executable by at least one processor of hardware processing circuit (s) of an apparatus. A hardware processing circuit includes at least one processor and at least one memory storing machine-readable instructions that are executable by the at least one processor of the hardware processing circuit. A processor includes any or some combination of an accelerator, a microprocessor, a core of a multi-core microprocessor, a microcontroller, a programmable integrated circuit, a programmable gate array, a digital signal processor, a central processing unit, a graphic processing unit, a tensor processing unit. A memory includes any or some combination of volatile or non-volatile memory (e.g., a flash memory, cache, a random-access memory (RAM) , and / or a read-only memory (ROM) ) . The memory stores the machine-readable instructions of the software and / or firmware for execution by the at least one processor of the hardware processing circuit. The machine-readable instructions are executable by the at least one processor of the hardware processing circuit to cause the hardware processing circuit to perform the actions or operations of the referee service described herein.
[0041] The referee service 105 may be a service provided by one of NFs of the communication network 125 (such as a gNB or an AMF) or may be a service provided by a separate NF (e.g., a new NF) . In some example embodiments, the referee service may be a service provided by an AI agent In some implementations, the referee service may be a type of management service provided by an operations, administration, and management system of a communication network. The referee service may assign one referee or more referees to an interworking group and may enforce a rule needed by an AI agents of an interworking group and may store actions generated by AI agents of an interworking group that are verified. In some example embodiments, the functionality of the referee service 105 may also be a network service. According to the rules, the referee service 105 may guarantee the that the actions of the AI agents in an interworking group are verified so that intentional or unintentional cheating does not occur. AI agent trustworthy issues may be avoided when AI agents are interworking or collaborating with each other on a task.
[0042] The referee service 105 may be provided by the communication network to the UEs 110 and 120. The actions or action generated by the AI agents on the UE 110 or 120 may be approved or rejected by the referee service 105. FIG. 2 illustrates a signaling flow of an example process 200 in accordance with some example embodiments of the present disclosure. The process 200 may involve the UE 110 or 120 and a referee service (e.g., referee service 105) . The AI agents on the UEs 110 and 120 may be part of a set of agents, and the UE 110 is a subscriber of a communication network. The UE 110 may take the responsibilities for the other AI agents in an interworking group, where the UE on which the AI agent is hosted has no subscription to the communication network.
[0043] As shown in FIG. 2, the UE 110 obtains (205) an action (referred to as a first action) from an AI agent (referred to as a first AI agent) hosted or deployed on the UE 110. The first action is generated by the first AI agent hosted on the UE 110 and the action is related to a task that a set of AI agents that includes at least the first AI agent, are interworking on (e.g., participating in) . The first action will be verified by the network through a referee service, as will be detailed in the following paragraphs.
[0044] The set of AI agents may be created by the UE 110. In some example embodiments, the UE 110 may create the set of AI agents, based on receipt of a trigger that indicates that a set of AI agents is to be created. The trigger may be from a user input or a user request or from a related application running on the UE 110. In an example, the user may select an application (e.g., a game application) running on the UE 110 and requiring collaboration of the UE 110 and other UEs (for example, the UE 120) . The selection of the user may be considered as a trigger for the creation of a set of AI agents on the UEs involved. If the UE 110 detects the selection of the user, the UE 110 may create the set of AI agents. In another example, starting or launching of the application may trigger creation of the set of AI agents. For example, in response to the starting or launching of the application, the UE 110 may create the set of AI agents.
[0045] The set of AI agents may be created or formed by the UEs based on a task and requirements. For example, the UEs 110 and 120 may communicate with each other to exchange IDs of their AI agents and the related set of rules. For instance, if the UEs are in proximity to each other, the UEs may communicate with each other to create a set or group of AI agents (e.g., an interworking group) over a sidelink of the UEs.
[0046] After the set of AI agents are created, the UE 110 may send, to the referee service 105, information related to the set of AI agents, for example. The information may comprise an identity (ID) of the set of AI agents, a set of rules for the task, and / or identities (IDs) of respective AI agents included in the set of AI agents. The ID of an AI agent on the UE 110 may be formed by a unique local name of the AI agent and an ID of the UE 110. After the referee service 105 receives the information related to the set of AI agents from the UE 110, the referee service 105 may send, to the UE 110, a confirmation of receipt of the information related to the set of AI agents.
[0047] In some example embodiments, the information related to the set of AI agents sent from the UE 110 to the referee service 105 may be used as a request for the referee service to verify and validate or approve actions related to the task generated by AI agents included in the set of AI agents.
[0048] By way of example, an AI agent on the UE 110 may be an LLM / LMM based agent and may initiate creation of an interworking group for a task. The interworking group may comprise a set of AI agents or group of AI agents that are participating or interworking together on a task and AI agent (e.g., the UE 100) may register the interworking group with the communication network. The interworking group may be identified by an identity or identifier (ID) of the group (referred to as group ID) . When the AI agent on the UE 110 registers the interworking group with the communication network, the AI agent on the UE provides information about the interworking group, including the group ID an identities or identifiers IDs of the AI agents (e.g., LLM / LMM based agents) of the AI agents in the interworking group, and also provides a set of rules for the task (for example, an application) . In some examples, the information about the interworking group may additionally comprise a type of the AI agent involved in the task, or if the type of AI agent involved in the task is not known to the communication network, knowledge triples (or their equivalent) that defines the task (e.g., the application) . Herein, the term “knowledge triple” may represent a piece of knowledge, or fact, or a result of an action, which consists of three components. It may be shared among different entities (e.g., AI agents) . A set of these “knowledge triples” may be used to identify a certain task (e.g., an application) . Optionally, the information about the interworking group may comprise a requested time duration for the AI agents to interwork together on the task. In this way, the UE 110 may request the network to establish a referee session for a specific interworking group and assign a referee to the interworking group.
[0049] Based on where the referee service 105 is located, the UE may send its requests via different protocols. For example, if the referee service is provided by a NF of a core network function, the UE may send a non-access stratum (NAS) message that includes the UE’s request to an AMF and an AMF of the core network may forward the UE’s requests to the NF of the core network. If the referee service 105 is as part of (e.g., provided by) a gNB, the UE 110 may send a radio resource protocol (RRC) protocol message that includes the UE’s request to the referee service 105. The UE 110 may also send its request via an application layer to an application function (AF) where the AF can discover available services of the communication network, including the referee service 105, directly from a NRF if the AF is a trusted AF, or from a NRF via a NEF if the AF is a non-trusted AF.
[0050] After the referee service 105 receives the request for the establishment of the referee session, the referee service 105 may establish a referee session. In some example embodiments, the referee service 105 may determine whether one or more UEs (such as the UEs 110 and 120) , which have AI agents that are in the set of AI agents (e.g., have AI agents that are in the interworking group) , have subscribed to the referee service 105. For example, the referee service 105 may check the subscriptions of UEs that have AI agents that in the set of AI agents determine whether the UEs have subscribed to the referee service. The subscriptions of the UEs that have AI agents in the set of may be stored in a UDM or UDR of the communication network. If the referee service 105 determines that the one or more UEs that have AI agents that in the set of AI agents have subscribed to the referee service, the referee service 105 may assign a referee service for referee session. In some example embodiments, the referee service 105 may assign a global identity (ID) to the set of AI agents, for example, which is to be used for communications with the referee service 105.
[0051] Upon the establishment of the referee session, the referee service may send, to the UE 110, a response to indicate the establishment of the referee session. For example, the referee service may send, to all the UEs having AI agents that are participating in the task, an address of the referee , a duration of the referee session (optional) , a set of rules, the global ID of the set of AI agents and possibly information about how an AI agent can terminate the referee session. In some examples, if the referee service rejects the request from the UE 110, the referee service may propose the UE 110, for example, changing the set of rules, or joining to one other set of AI agents (e.g., another interworking group) which is already being refereed by the referee service using the requested set of rules.
[0052] In some other example embodiments, the set of AI agents may be created by another UE (such as the UE 120) having an AI agent participating in the task. In this case, the UE 120, on behalf of other UEs having AI agents that are participating in the task, may send the information related to the set of AI agents, to the referee service 105, for example, to request for establishment of a referee session.
[0053] In some example embodiments, the establishment of a referee session may take into account consent from other UEs having AI agents that are participating in the task. For example, in the example embodiments wherein the set of AI agents are created by the UE 120, the referee service 105 may send, to the UE 110 (and other related UEs) , a request for a related AI agent to be included in the set of AI agents participating in the task. After the UE 110 receives the request, the UE 110 may send, to the referee service 105, a response which indicates consent of the AI agent to be included in the set of AI agents participating in the task (e.g., to be included in the interworking group) .
[0054] For example, the referee service 105 may send a request with the information related to the set of AI agents to other UEs (or even each of the UEs) and indicate the identity (ID) of the AI agent of the requesting UE. Each of the AI agents or UEs may send a response to the request, where the response indicates the UE’s consent for its AI agent to join the set of AI agents. If one of the UEs declines the request to join the set of AI agents, the referee service may reject the creation of the set of AI agents.
[0055] In this way, the referee service 105 may request the other UEs confirm whether they are in the set of AI agents or not. After receipt of the confirmation from all other UEs, the referee service 105 may assign a global ID to the set of AI agents and start a referee session for the set of AI agents and take referee actions according to the set of rules provided by the initiating UE (for example, the UE 110) . Herein, the term “referee session” and “service session” used are interchangeably used to refer to a “referee service” that is established by a network function (e.g., a network referee) . The establishment of a referee session may comprise the whole process from the creation of the set of AI agents to completion of the referee service for the set of AI agents.
[0056] In some examples, each of the agents or UEs may respond with its signature for the proposed set of AI agents, if it consents to join the set of AI agents, for example, for interworking for the task such as an application. The referee service 105 may verify the signatures from the UEs. Alternatively, if the application is not known before, the referee service 105 may verify the set of rules and the knowledge triples (or their equivalent) defining the application involved in the task. If the signatures are authentic and the set of rules and the knowledge triples (or their equivalent) are consistent and legitimate, the referee service 105 may announce the creation of the proposed set of AI agents (for example, identifiable with an ID of the set of AI agents) to the UEs and, possibly to other relevant network entities. The referee service 105 may record the application and the set of its rules to a knowledge base, if the application is not known before. If a signature of one of the UEs is unauthentic or the set of rules or the knowledge triples are illegitimate, the referee service 105 may reject the registration and creation of the set of AI agents.
[0057] Alternatively, the set of AI agents may be created by the referee service 105 of the communication network. In some example embodiments, the referee service 105 may create the set of AI agents, based on receipt of a trigger that indicates the set of AI agents is to be created. For example, in some cases (e.g., traffic law enforcement) , the set of AI agents may be created directly by the referee service 105 without the initiation and consent from any UE participating in the task. The trigger to create the set of AI agents may be from a UE. For example, the activation of an AI agent on the UE 110 or 120 for a related application requiring collaboration of multiple AI agents may trigger the creation of the set of AI agents by the referee service 105. Alternatively, or in addition, the trigger may be from a monitoring function of the referee service 105. For example, in a use case of collaboration among self-driving cars, the referee service 105 may monitor all the vehicles or the AI agents on a road of interest. If some AI agents on the UEs are monitored to enter a certain geolocation, the referee service 105 may create a set of AI agents.
[0058] The referee service 105 may create a set of AI agents according to the conditions of the UEs having the AI agents. For example, the conditions of the UEs may be membership of the UEs, proximity of the UEs to each other, a type of subscriptions of the UEs, and a set of rules for the task.
[0059] After the set of AI agents are created by the referee service 105, the referee service 105 may send, to the UE 110, a request for the AI agent to be included in the set of AI agents participating in the task. After the UE 110 receives this request, the UE 110 may send, to the referee service 105, a response which indicates consent of the AI agent to be included in the set of AI agents participating in the task.
[0060] Alternatively, or in addition, a UE may automatically request for its AI agent to join an existing set of AI agents or a set of AI agents that is in the process of being created. For example, the UE 110 may send, to the referee service 105, a request for the AI agent to join the set of AI agents participating in the task. After the referee receives this request, the referee service may send a response which indicates approval of the request. In some cases (e.g., law enforcement) , a UE may not need to make a request for its AI agent to join a set of AI agents. Rather, the referee service 105 may include a relevant AI agent of a UE in the created set of AI agents, even without the initiation of the UE.
[0061] In some example embodiments, the capability of the referee service 105 for establishing and providing a referee session may be exposed to UEs. For example, the referee service 105 may send, to one or more UEs associated with the set of AI agents (including the UE 110) , capability information of the referee service 105, which may include information indicating that the referee service 105 is capable of establishment of a referee and information about capabilities of the referee service 105 or referees the referee service 105 can establish. Based on the capability information that is received, the UE 110 may request for the AI agent to join an appropriate set of AI agents.
[0062] For example, in an autonomous driving use case, the referee service 105 may be provided by a gNB covering one area and then the gNB may broadcast the capability information of the referee service 105 to the UEs. Alternatively, a communication network may expose its services (including the referee service) and its capabilities to UEs, via universal integrated circuit card (UICC) codes or in a short message service (SMS) message sent by the communication network to the UEs.
[0063] Alternatively, a set of AI agents or a group of AI agents may be already created and operational. In this case, a referee implemented by the referee service 105 may be assigned to the created and operational set of AI agents. The assignment of a referee may be requested by any participant in a set of AI agents or a group of AI agents. The assignment of a referee may be requested in the case where the referee service 105 initiates the creation of a set of AI agents or a group of AI agents and assigns a referee to the set of AI agents or the group of AI agents posteriori. Alternatively, or in addition, the assignment of a referee may be requested in the case where referee handover occurs, for example, when some UEs (or agents) have subscribed to communication networks of different operators, or subscribed to different management domains of the same operator. In this case, two or more referees may be assigned to the set of AI agents, and one of the referees may serve as a main referee while others may serve as alternatives. If referee responsibility of the main referee is handed over to an alternative referee, the assignment of the new main referee may be needed.
[0064] In some example embodiments, the referee service 105 may be discoverable by UEs communicating with the communication network. Table 1 shows an example data (information) model of the referee service 105. Table 1
[0065] Every time a new set of AI agents or a group of AI agents are created, an instance of the attribute referred to as GroupInfoItem is generated and added to the attribute referred to as a list activeGroupList. It is removed from the list when the group is not needed anymore. The parameter “P” represents a present modifier, indicating whether an attribute is present in the message or not, where “M” indicates mandatory, “CM” indicates conditionally mandatory, “O” indicates optional, and “CO” indicates conditionally optional.
[0066] Table 2 shows a definition of a data model for the attribute GroupInfoItem where the detailed information about the corresponding group is defined. Table 2
[0067] Table 3 shows a definition of a data model for a referee service consumer to request the referee service, referred to as “RefereeRequest” . Table 3
[0068] In one example, the OAM can configure a PCF with the policy rules related to different areas and the referee service.
[0069] After a session for a referee service is established and a referee is assigned to a set of AI agents or a group of AI agents, all AI agents (e.g., UEs on which the AI agents are hosted) may have access to the referee and may communicate with the referee. For example, an AI agent (e.g., a LLM / LMM-based agent) on any of UEs may start interworking with other AI agents on a task (e.g., may participate in the task) .
[0070] Still with refence to FIG. 2, the UE 110 sends (210) , to the referee service, a request for verification of the first action generated by the AI agent on the UE 110, where the first action is related to the task that the set of AI agents or group of AI agents are participating in. The request comprises information related to the first action which the referee of the referee service 105 uses to verify the first action. This information may specify the first action, context of the UE 110 associated with the first action, and / or the task. For example, in the case where the UE 110 is in road traffic, the UE 110 may sends information related the first action as well as the context (e.g., speed, a vehicle status, a location, etc. ) to the referee. In some examples, when the referee service 105 is exposed to UEs, this exposure may serve as a trigger for a corresponding agent at the UE to start reporting its relevant action, the context of the action and / or the task to the referee service, for example, if it is preconfigured for the UE by its user or law enforcement.
[0071] Table 4 shows definition of a data model for a consumer of the referee service to request verification of an action, referred to as “ActionValidationRequest” . Table 4
[0072] The data model (or information model) may define semantics of a message between entities. For example, in Table 4, the data model defines 4 entries in the message (where one, indicated as “O” , is optional) , with types of the entries.
[0073] After the referee service receives (215) the request, the referee service (or the referee established by the referee service) performs (220) the verification of the first action. The verification may be based on the set of rules defined for the task. Upon the verification of the first task, the referee service 105105 sends (225) a response to the UE 110 to indicate a result of the verification. Correspondingly, the UE 110 receives (230) the response. Any member UE may send its intended action to the referee (otherwise referred to as a network referee) for validation or stamping of an action generated by its AI agent to determine whether the action complies with the defined set of rules for a task that the AI agent is participating in.
[0074] In some example embodiments, the referee service 105 may verify whether the first action is approved or rejected. In this example, the UE 110 may need to wait for a response from the referee service 105 before acting. For example, if the response indicates the approval of the first action, the UE 110 may perform the first action related to the task. If the response indicates the rejection of the first action, the UE 110 may obtain, from the AI agent, a further action (referred to as a second action) related to the task, and sending, to the referee service 105, a request comprising information related to the second action, to further verify whether the second action is approved or rejected.
[0075] For example, the referee service 105 may evaluate the request from the UE 110 based on the defined set of rules and send either approval or rejection to the UE 110. The UE 110 (e.g., the AI agent on the UE 110) may act accordingly-if approved, it will proceed with the intended action; if rejected, it will adjust its action and request validation again.
[0076] In some example embodiments, there may be a time threshold (e.g., a predetermined period of time) that the UE 110 must receive the response from the referee service 105. If the UE 110 does not receive a response from the referee service within the time threshold (e.g., the predetermined period of time) , the UE 110 may re-send the request.
[0077] In some example embodiments, the UEs may not need to wait to receive a response from the referee service that indicates approval from the network referee for their action. For example, in traffic law monitoring cases, the UE 110 may execute the action (e.g., the AI agent may cause the UE to perform the action) before the UE 110 reports this action to the referee service 105. As the referee service 105 may already know the set to AI agents, the referee service 105 (or the referee assigned by the referee service 105 to the set or group of AI agents) may monitor the related UE (e.g., the UE 110) and its actions (and context) to detect whether the action violates the rules for the task. If the referee service 105 (or the referee) detects an action that is performed by a UE that violates the rules for the task, the referee service 105 may send a warning and possible correct actions to the UE (e.g., the AI agent on the UE) and even the other UEs (e.g., the AI agents on the other UEs) . In some examples, the referee service 105 may store activities, intended actions, and the corresponding validations of actions generated by AI agents on UEs for future use, such as auditing, regulatory compliance, or providing historical results to UEs (e.g., AI agents on UEs) upon request.
[0078] By way of example, after a set of AI agents or a group of AI agents is determined to participate in a task, the AI agent (e.g., a LLM / LMM-based agent) on a UE may send information that specifies an action generated by the AI agent and a context related to the action) to another AI agent (e.g., LLM / LMM-based agent) on another UE. This information may be first sent to the referee service 105 . For some tasks for a set of AI agents or a group of AI agents, the referee service 105 (or the referee) may request the AI agents send their information to the referee service 105. For example, all information exchanged between AI agents in a set of AI agents or a group of AI agents may need to be sent first to the referee service 105 (or the referee) for verification. For some other tasks, the referee may only check selected pieces of data, e.g., only actions decided by a selected agent in the set of AI agents are first sent to the referee for verification. In some other cases such as the auto-driving in road traffic, the agent may take an action before it reports the action to the referee. The referee then may react to the already taken action accordingly.
[0079] The referee service 105 may verify the data by reasoning over the set of rules and the knowledge triples (or their equivalent) for the interworking, with the knowledge base. The verification may possibly also utilize the context-based recommendation (s) of other network-side LLM / LMM agent (s) . If the data complies with the set of rules and the knowledge triples (or their equivalent) , the referee may make a signature on the data and then, if needed, forward the data and the signature to the other LLM / LMM-based agent on the other UE. Moreover, the referee may add the forwarded data to a new step of the interworking and the interworking steps (from beginning until now) are kept into the knowledge base.
[0080] If the data does not comply with the set of rules and the knowledge triples (or their equivalent) , the referee may inform the original sending agent that data does not comply with the set of rules and the knowledge triples (or their equivalent) . In addition, if the “not-complying” occurs many times (for example, the number of occurrences exceeds a threshold) , the referee can take an action (e.g., warning, blocking, taking-over the control of, etc. ) the sending agent and its UE or, if relevant, may disband the whole interworking group. In addition, in the case of traffic law enforcement, the “not complying” incident of the agent and its UE may be reported to the law enforcement agency if it violates the traffic law. During the interworking, any of the agents in the set of AI agents may request for the recorded interworking steps of the set of AI agents from the referee.
[0081] In some examples, more than one referee may be assigned to a set of AI agents. For example, UEs subscribed to communication networks of different operators may be associated with the same group of AI agents for road traffic enforcement. In this case, the referee services of different communication networks may need to decide together or the decision of one referee service needs to be verified by another referee serving. In some example embodiments, the verification of the first action generated by an AI agent on the UE 110 is based on coordination between the referee service of one communication network and the referee service of another communication network . Both the referee services are configured to provide verification of actions generated by the set of AI agents. In some example embodiments, the coordination between the referee services may be controlled by a further referee service that is acting as a referee service broker. Some example embodiments in this regard will be described in the following paragraphs with reference to FIG. 5.
[0082] The referee session established by the referee service 105 may be terminated, for example, responsive to a request from a UE. For example, in some example embodiments, the UE 110 may send, to the referee service 105 a request for the referee session to be terminated, based on receipt of a trigger that causes the sending of the request. The trigger may be a user input to request the termination of a referee session, closing of an application related to the task, accomplishment of the task, leaving of the UE from a certain area, and / or the like. The request may include information indicative of a reason for the termination of the referee session (for example, the trigger) . After the referee service receives this request, the referee service may terminate the referee session and send, to the UE 110, an indication that the referee session has been terminated.
[0083] The request for the termination of the referee session may be an explicit or implicit request. For example, the UE 110 may explicitly request termination of a referee session. Alternatively, the UE 110 may request to leave the set of AI agents. Then, the referee session for the set of AI agents or group of AI agents may be terminated and all AI agents (e.g., UEs which the AI agents are on) of the set of AI agents or group of AI agents may be notified that the referee session has been terminated.
[0084] In addition to the request from the UE, or as an alternative, the referee session may be terminated upon expiration of duration of the referee session. Alternatively, or in addition, the referee service 105 or the referee may decide to terminate the referee session or disband the set of AI agents by itself, for example, if the actions generated by the AI agents are determined by the referee to violate the rules for the task many times, as described above. By way of example, the referee may decide to disband the set of AI agents or group of AI agents if one of following situations occurs: the interworking of the AI agents in the set of AI agents or group of AI agent on a task is considered to have been ended by the referee; the initiating AI agent (e.g., LLM / LMM based agent) on a UE requests the disbanding of the set of AI agents or group of AI agents and other AI agents in the set of AI agents or group of AI agents agree; and interworking of the AI agents in the set of AI agents or group of AI agent on a tahas achieved its goal. In some examples, an identity (ID) of the set of AI agents, the set of rules and the knowledge triples (or their equivalent) , and the interworking steps may be stored or kept for reference and verification even after the referee service has been terminated.
[0085] Example process that involves the referee service 105, the UEs 110, 120, in which a UE 110 and the referee service 105 create the set of AI agents will be described below with reference to FIGS. 3 and 4, respectively. Reference is first made to FIG. 3 where the set of AI agents or group of AI agents are created by UEs. In an example process 300, a set of AI agents (e.g., a group of AI agents) may be created by UEs (such as the UEs 110 and 120) , and then the UE 110 on behalf of other UEs may send an identifier of the set of AI agents of the group of AI agents, a set of rules (also referred to as a rule set) , and an identifier of each AI agent in set of AI agents or group of AI agents (or identifiers of the UEs on which the AI agents are on) to the referee service 105. The referee service 105 may request other AI agents (e.g., other UEs) in the set or group of AI agents to confirm whether AI agent is included in the set of AI agents or group of AI agents or not. After the confirmation is received from all AI agents in the set of AI agents or group of AI agents (e.g., received from UEs of the AI agents) , the referee service 105 may assign a global group identity (ID) (also referred to as a referee group identity (group ID) to the set of AI agents of the group of AI agents and then establish a referee session for the set of AI agents or the group of AI agents according to the group rule provided by the initiating UE 110. The group and the referee session may end when one member in the group indicates the quit (with a result) of the group. During the referee session, the referee service 105 may act according to the defined principle (s) .
[0086] In this example, all UEs may have such service subscriptions with the referee service 105 or its network operator. Each UE may send an intended action to the referee service 105 and then approval / rejection may be feedbacked to the UE before the UE may take the action (if approval) . The referee service 105 may store the UE steps in sequence for future use (e.g., notifying the member agents in the group when relevant) .
[0087] In operations, as shown in FIG. 3, at 310, a related application may start in the UEs 110 and 120 or the UEs 110 and 120 may be in the same area or in proximity, which triggers creation of a group of agents on the UEs (where such agents may also be referred to as UE agents) . At 312, the group may be created by the UEs 110 and 120. One UE (for example, the UE 110) may be selected to send a local group ID, the rule set of the task and IDs of group members. At 314, the UE 110 may send the local group ID, the rule set of the task and the IDs of group members to the referee service 105, to request for a referee service. The request from the UE 110 to the referee service 105 may be based on any protocol and solutions which are standardized in 3GPP to access a new service which the network can provide to the UE.
[0088] In some examples, different protocols may be used depending on the location of the referee service 105. For example, if the referee service 105 is at a part of one NF functionality in a core network, the UE 110 may send its request via an AMF 304 to that NF. If the referee service is at a gNB 302, the UE 110 may send its request via a RRC protocol to the referee service 105.
[0089] At 316, the referee service 105 may check the subscriptions of UEs (from a UDM / UDR 308) for the requested service. At 318, the referee service105 may send a confirmation request about UE1 request, to other UEs such as the UE 120. In this way, the referee service105 may ask other related UEs to confirm whether their agents are in this group or not. After the confirmation from all UEs, at 320, the referee service 105 may assign a global group ID to the group and start the referee session for the group. The referee service 105 may take referee actions according to the rule set of the group provided by the initiating UE 110.
[0090] At 322, referee session establishment may be notified to the group. At 324, one the UEs (for example, the UE 120) may select an intended action. At 326, the UE 120 may request the referee service 105 to stamp the intended action. At 328, the referee service 105 may check with rule sets and verify the intended action. The referee service 105 may evaluate the request based on the defined rule set maintained at a PCF 306. At 330, the referee service 105 may send approval or rejection based on rules. At 332, the UE 120 may act accordingly. If the action is approved, the UE 120 may proceed with the intended action. If the action is rejected, the UE 120 may adjust its action and request validation again. Any UE 110 or 120 may send its intended action to the referee service 105 for validation to determine whether it complies with the defined rule set. The UE may need to wait for a response from the referee service 105 before acting. At 334, the referee service 105 may store the UE steps (for example, the intended actions of the UEs and their corresponding validation results) in sequence for future use.
[0091] The group and the referee session may end if one member in the group indicates the quit (with a result) of the group. At 336, the UE 120 may close the applications or leave the area. At 338, the UE 120 may request to quit the referee session with a reason. At 340, the referee service 105 may terminate the referee service. At 342, the referee service 105 may inform the UEs 110 and 120 about service end and disband the group. Alternatively, when interworking has achieved its goal, the referee service 105 may notify to all UEs that the service session ends. The referee service may archive the action steps of the group members, its group ID, and the associated rule set for future use.
[0092] FIG. 4 illustrates an example process 400 of a referee service where the set of AI agents are created by the network. For example, in one scenario, the referee service 105 may know the rules of the task in advance, for example, the traffic enforcement rules for vehicles in proximity. The referee service 105 may ask the UEs (when they are in road traffic) to report its action and context (e.g., speed, vehicle status, locations, etc. ) to the referee service 105. The referee service 105 may monitor the actions and context of the UEs. The referee service 105 may only send a warning to the UE and its related UEs (e.g., when they are in certain proximity) if the referee service 105 notices something violations of the rules. The referee service 105 may store the actions and contexts in sequence. It may also forward the records to a referee broker, for example, which may manage a certain geo-area. In another scenario, the referee service 105 may not know about rules initially. The UEs may request the referee service along with the rules. Then, the network may form the agent groups and configure a referee service (for example, the referee service 105) with the UE-provided rule set. Groups (including its rule set) may be created or managed dynamically by the referee service 105, where the agents on the UEs may join or leave a group based on the decisions of the referee service 105. In this example, UEs do not participate in the group creation and management. A UE may take an action without the need to wait for the approval from the referee service 105.
[0093] In operations, as shown in FIG. 4, at 450, the referee service 105 may trigger creating or assignment of a group of AI agents, e.g., if an agent on a UE is activated for a relevant application, a UE is entering certain context, e.g., certain geolocation) , etc. At 452, the group may be created by referee service 105, or the referee service 105 may create a group on demand e g. based on UE proximity, UE subscriptions, a rule set of a task. For example, the UEs may have different applications, and for example, launch one of the applications. An SMF 445 may know about an ID of the application, and thus the referee service 105 may obtain the ID of the application from the SMF 445. The referee service 105 may also check the locations of the UEs, for example, from an LMF 447. After that, the referee service 105 may decide that the UEs running this application in this area may be grouped together. At 454, the referee service 105 may expose the registered referee services and the corresponding capabilities to UEs (for example, the UEs 110 and 120) .
[0094] Based on the received information about the service availability, a UE may request for its agent to join an existing group or one that is in the process of being established. At 456, the UE 110 may identify the need for the referee service 105, based on the running applications or its location. At 458, the UE 110 may send a request for its agent to join the group. At 460, the referee service 105 may check subscriptions of the UE 110 from the UDM / UDR 308. At 462, the referee service 105 may approve the joining request.
[0095] Alternatively, in some cases (e.g., law enforcement) , a UE may not need to make the request to join. As shown in FIG. 4, at 464, the referee service 105 may identify the need for the referee service based on running applications or locations of the UEs. At 466, the referee service 105 may check subscriptions of the UEs. At 468, the referee service 105 may request the UEs 110 and 120 to join. At 470, the UEs 110 and 120 may give consent to join.
[0096] At 472, the referee service 105 may configure the group to send actions and contexts in specific intervals. At 474, the UEs 110 and 120 may monitor the contexts, select actions and store them to send. At 476, the UE 110 may send its actions and context to the referee service 105. At 478, the referee service 105 may start to monitor the actions and contexts of the UE 110. At 480, violation detection may be performed based on the monitoring by the referee service 105. At 482, if such a violation is detected, the referee service 105 may send the warning to all UEs. At 484, the referee service 105 may store actions and contexts of the UEs for future use.
[0097] In some examples, a referee services of communication networks of different operators may be used by different UE having AI agents of an interworking group (e.g., a set or group of AI agents) , as described above. FIG. 5 illustrates an example process 500 of coordination between referee service of different communication networks in accordance with some example embodiments. The involvement of two or more referee services of different communication networks is because the UEs have their subscriptions for a referee service of the communication networks of different network operators. UEs with subscriptions for a referee service of communications networks of different network operators may be reporting to and monitored by the referee service of the communication networks of their respective operators. For example, UE1 may have a subscription for a referee service of a communication network of operator A and UE2 may have a subscription for a referee service of a communication network of operator B. A referee service 530 (referred to as referee service A) offered by operator A may observe UE1, and a referee service 540 (referred to as referee service B) offered by operator B may observe UE2. UE 1 may create an interworking group (e.g., a set or group of AI agents) , and UE2 may create another interworking group (e.g., another set or group of AI agents) . The two separated interworking groups (e.g., two sets or groups of AI agents) created by UE1 and UE2 may be managed by the referee services A and B, respectively. Alternatively, the separated local referee groups may be created by the networks separately. The separated local groups may join and form a global group of AI agents together, by the coordination between the referees at their networks. The referee services of the communications networks of operators A and B may be coordinated. The operations of the two referee services A and B are as described above, while the coordination between the referee services of the different communication networks may be handled as described above and is transparent to the UEs.
[0098] In operations, as shown in FIG. 5, at 560, an agent 510 of UE1 and an agent 520 of UE2 of an interworking group. In an example 570, referee service A and referee service B may directly work together as a logically central referee. At 572, a referee service may be provided from referee service A to UE1. At 574, a referee service may be provided from referee service B to UE2. At 573, coordination may be performed between the two referee services. For example, the referees (e.g., referees A and B) may share their observations / requests of the UEs (e.g., UE1 of referee service A and UE2 of referee service B) in real-time when the agents 510 and 520 of the UEs are inside the same global group (e.g., in a same game or in certain proximity of road traffic) . The two referees may together act as a logically centralized referee for the UEs (e.g., UE 1 and UE2) and use the refereeing steps as described above. Likewise, the two referees may store the actions and contexts in sequence and may forward their records to a cross-operator broker 540.
[0099] In another example 575, referee service A and referee service B may use the cross-operator broker 540 to act as a logically central referee. At 576, a referee service may be provided from referee service A to UE1. At 578, a referee service may be provided from referee service B to UE2. At 577, the cross-operator broker 540 may act as the central referee to coordinate the two referees. In this example, the referees (e.g., referee service A and referee service B) may not act as a logically centralized referee for the UEs (e.g., UE1 and UE2) themselves. Rather, they may send their observations or requests of the UEs to the broker 540. Then, the broker 540 may integrate the information and check if it is related (e.g., in the same game or in certain proximity) . If yes, the broker 540 may act as the global referee and dynamically create the global referee group / manage the global group (e.g., for UE1 and UE2) by using the referee steps defined above. The local referees (e.g., referee service A and referee service B) may act as the delegates / proxies of the global referee. A local referee (e.g., referee service A) may send the local warning to its UE if this local referee can determine its UE (e.g., UE1) has already violated the rule.
[0100] FIG. 6 shows a flowchart of an example method 600 performed or carried out by a UE, such as the UE110 or 120 of FIG. 1.
[0101] At block 610, the UE 110 obtains, from an AI agent on the user equipment, a first action generated by the AI agent. The action is related to a task that a set of AI agents are participating in. The set of AI agents include at least the AI agent. At block 620, the UE 110 sends, to the referee service 105 of a communication network, a request for verification of the first action. The request comprises information related to the first action. The request causes a referee service 105 to verify the first action. In some example embodiments, the information related to the first action may comprise a context of the user equipment and the task associated with the first action. At block 630, the UE 110 receives a response to the request, where the response indicates a result of the verification.
[0102] In some example embodiments, the response may indicate approval of the first action. In these example embodiments, the UE 110 may perform the first action related to the task. In some example embodiments, the response may indicate rejection of the first action. In these example embodiments, the UE 110 may obtain, from the AI agent, a second action related to the task. The UE 110 may send, to the referee service 105, a request for verification of the second action. The request may comprise information related to the second action, where the request causes the referee service 105 to verify whether the second action is approved or rejected.
[0103] In some example embodiments, the UE 110 may create the set of AI agents, based on receipt of a trigger that indicates the set of AI agents is to be created. The UE 110 may send, to the referee service 105, information related to the set of AI agents. The information may comprise an identity of the set of AI agents, a set of rules for the task, and identities of respective AI agents included in the set of AI agents. The UE 110 may receive, from the referee service 105, a confirmation of receipt of the information related to the set of AI agents.
[0104] In some example embodiments, the UE 110 may send, to the referee service 105, a request for establishment of the referee service for verification of actions related to the task generated by AI agents included in the set of AI agents. The UE 110 may receive, from the referee service 105, a response to the request, the response indicating the establishment of the referee service.
[0105] In some example embodiments, the UE 110 may receive, from the referee service 105, a request for the AI agent to be included in the set of AI agents participating in the task. in response to receiving the request, the UE 110 may send, to the referee service 105, a response to the request. The response indicates consent of the AI agent to be included in the set of AI agents participating in the task.
[0106] In some example embodiments, the UE 110 may send, to the referee service 105, a request for the AI agent to join the set of AI agents participating in the task. The UE 110 may receive, from the referee service 105, a response to the request, the response indicating approval of the request.
[0107] In some example embodiments, the UE 110 may receive, from the referee service 105 prior to the obtaining the first action, capability information of the NF 130. The capability information includes information indicating that the referee service 105 is capable of establishment of the referee service and information about capabilities of the referee service.
[0108] In some example embodiments, the UE 110 may send, to the referee service 105, a request for the referee service to be terminated, based on receipt of a trigger that causes the sending of the request. The UE 110 may receive, from the referee service 105, an indication that the referee service has been terminated. In some example embodiments, the request may include information indicative of a reason for the termination of the referee service.
[0109] FIG. 7 shows a flowchart of an example method 700 of a network function providing a referee service in accordance with some example embodiments. The referee service is being used by a set or group of UEs interworking (e.g., participating) in a task. The referee service may be or may be implemented as a referee entity as described above. For the purpose of discussion, the method 700 will be described from the perspective of the referee service 105 with reference to FIG. 1.
[0110] At block 710, the referee service 105 receives, from the UE 110, a request for verification of a first action of the UE 110. The request comprises information related to the first action. The first action is related to a task that a set of AI agents are participating in, and the set of AI agents include at least an AI agent on the UE 110. In some example embodiments, the information related to the first action may comprise a context of the UE 110 and the task associated with the first action. At block 720, in response to receiving the request, the referee service 105 performs a referee service to verify the first action. At block 730, the referee service 105sends, to the UE 110, a response to the request, wherein the response indicates a result of the verification.
[0111] In some example embodiments, the referee service 105 may receive, from the UE 110, information related to the set of AI agents, the information comprising an identity of the set of AI agents, a set of rules for the task, and identities of respective AI agents included in the set of AI agents. The referee service 105 may send, to the UE 110, a confirmation of receipt of the information related to the set of AI agents. The referee service 105 may also send, to the UE 110, at least one of an accepted duration of the referee service, a global group ID of the set of AI agents, or how to terminate the referee service.
[0112] In some example embodiments, the referee service 105may receive, from the UE 110, a request for establishment of the referee service for verification of actions related to the task generated by AI agents included in the set of AI agents. in response to receiving the request, the referee service 105 may establish the referee service. The referee service 105 may send, to the UE 110, a response to the request, the response indicating the establishment of the referee service. Alternatively, if the request is rejected, the response may indicate the reason of the rejection or possible solutions and recommendations provided from the network function.
[0113] In some example embodiments, the referee service 105 may determine whether one or more UEs associated with the set of AI agents have subscribed to the referee service. In accordance with a determination that the one or more UE 110s have subscribed to the referee service, the referee service 105 may establish the referee service. In some example embodiments, the referee service 105 may assign a global identity to the set of AI agents.
[0114] In some example embodiments, the referee service 105 may create the set of AI agents, based on receipt of a trigger that indicates the set of AI agents is to be created. The referee service 105 may send, to the UE 110, a request for the AI agent to be included in the set of AI agents participating in the task, and receive, from the UE 110, a response to the request. The response may indicate consent of the AI agent to be included in the set of AI agents participating in the task.
[0115] In some example embodiments, the referee service 105 may receive, from the UE 110, a request for the AI agent to join the set of AI agents participating in the task. In response to receiving the request, the referee service 105 may send, to the UE 110, a response to the request, the response indicating at least one of: approval of the request, a duration of the approval, or a duration of availability of the set of AI agents.
[0116] In some example embodiments, the referee service 105 may send, to one or more UE 110s associated with the set of AI agents, capability information of the network function, the capability information including information indicating that the network function is capable of establishment of the referee service and information about capabilities of the referee service.
[0117] In some example embodiments, the referee service 105may receive, from the UE 110, a request for the referee service to be terminated. In some example embodiments, the request includes information indicative of a reason for the termination of the referee service. I n response to receiving the request, the referee service 105 may terminate the referee service, and send, to the UE 110, an indication that the referee service has been terminated.
[0118] In some example embodiments, the verification of the first action may be based on coordination between the network function and another network function of the communication network, where both the network functions are configured to provide the referee service for the verification of actions generated by the set of AI agents. In some example embodiments, the coordination between the network functions may be controlled by a further network function.
[0119] In some example embodiments, apparatuses capable of performing any of the methods 600 and 700 may comprise means for performing the respective operations of the methods 600 and 700 and any of the embodiments thereof. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module. The apparatuses may be implemented as or included in the UE 110 and the referee service 105 in FIG. 1.
[0120] FIG. 8 is a simplified block diagram of a device 800 that is suitable for implementing example embodiments of the present disclosure. The device 800 may be a user equipment, for example, the user equipment 110 or the user equipment 120 or an apparatus for a telecommunication network that implements the referee service 105 disclosed herein. As shown, the device 800 includes one or more processors 810, one or more memories 820 coupled to the processor 810, and one or more communication modules 840 coupled to the processor 810.
[0121] The communication module 840 is configured to enable bidirectional communications. The communication module 840 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 devices. In some example embodiments, the device is a UE and the communication module 840 may include at least one antenna or an antenna array.
[0122] The one or more processors 810 may include one or more of the following: a central processing unit, a tensor processing unit, a graphics processing unit, a hardware accelerator, microprocessors, digital signal processors (DSPs) and a processor based on multicore processor architecture, an application specific integrated circuit , or a field programmable gate array, as non-limiting examples.
[0123] The memory 820 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) 824, 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) 822 and other volatile memories that will not last in the power-down duration.
[0124] A computer program 830 includes computer executable instructions that are executed by the associated processor 810. The instructions of the program 830 may include instructions for performing operations / acts of some example embodiments of the present disclosure. The program 830 may be stored in the memory, e.g., the ROM 824. The processor 810 may perform any suitable actions and processing by loading the program 830 into the RAM 822.
[0125] The example embodiments of referee service of the present disclosure may be implemented as a program 830 so that when instructions of the program are executed by the one or more processors 810 of the device 800, the device 900 is caused to perform any process of the referee service as disclosed herein.
[0126] In some example embodiments, the program 830 may be tangibly contained in a computer readable medium which may be included in the device 800 (such as in the memory 820) or other storage devices that are accessible by the device 800. The device 800 may load the program 830 from the computer readable medium to the RAM 822 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) .
[0127] 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.
[0128] 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.
[0129] 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.
[0130] 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.
[0131] 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.
[0132] 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 user equipment comprising:at least one processor; andat least one memory storing instructions of a client, wherein the instructions, when executed by the at least one processor, causing the user equipment to perform operations, the operations comprising:obtaining, from an artificial intelligence (AI) agent on the user equipment, a first action generated by the AI agent, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least the AI agent;sending, to a network function of a communication network, a request for verification of the first action, the request comprising information related to the first action, wherein the request causes a referee service of the network function to verify the first action; andreceiving a response to the request, wherein the response indicates a result of the verification.2.The user equipment of claim 1, wherein the response indicates approval of the first action and wherein the operations further comprise:performing the first action related to the task.3.The user equipment of claim 1, wherein the response indicates rejection of the first action and wherein the operations further comprise:obtaining, from the AI agent, a second action related to the task; andsending, to the network function, a request for verification of the second action, the request comprising information related to the second action, wherein the request causes the network function to verify whether the second action is approved or rejected.4.The user equipment of any of claims 1 to 3, wherein the operations further comprise:creating the set of AI agents, based on receipt of a trigger that indicates the set of AI agents is to be created;sending, to the network function, information related to the set of AI agents, the information comprising an identity of the set of AI agents, a set of rules for the task, and identities of respective AI agents included in the set of AI agents; andreceiving, from the network function, a confirmation of receipt of the information related to the set of AI agents, and at least one of an accepted duration of the referee service, a global group ID of the set of AI agents, or how to terminate the referee service.5.The user equipment of any of claims 1 to 3, wherein the operations comprise:sending, to the network function, a request for establishment of the referee service for verification of actions related to the task generated by AI agents included in the set of AI agents; andreceiving, from the network function, a response to the request, the response indicating the establishment of the referee service, or if the request is rejected, the reason of the rejection or possible solutions and recommendations provided from the network function.6.The user equipment of any of claims 1 to 3, wherein the operations further comprise:receiving, from the network function, a request for the AI agent to be included in the set of AI agents participating in the task; andin response to receiving the request, sending, to the network function, a response to the request, the response indicating consent of the AI agent to be included in the set of AI agents participating in the task.7.The user equipment of any of claims 1 to 3, wherein the operations further comprise:sending, to the network function, a request for the AI agent to join the set of AI agents participating in the task; andreceiving, from the network function, a response to the request, the response indicating at least one of approval of the request, a duration of the approval, or a duration of availability of the set of AI agents.8.The user equipment of any of claims 1 to 7, wherein the operations further comprise:receiving, from the network function prior to the obtaining, capability information of the network function, the capability information including information indicating that the network function is capable of establishment of the referee service and information about capabilities of the referee service.9.The user equipment of any of claims 1 to 8, wherein the information related to the first action comprises a context of the user equipment and the task associated with the first action.10.The user equipment of any of claims 1 to 9, wherein the operations further comprise:sending, to the network function, a request for the referee service to be terminated, based on receipt of a trigger that causes the sending of the request; andreceiving, from the network function, an indication that the referee service has been terminated.11.The user equipment of claim 10, wherein the request includes information indicative of a reason for the termination of the referee service.12.An apparatus for a communication network, the apparatus comprising:at least one processor; andat least one memory storing instructions of a network function of the communication network, the network function configured for establishing and providing a referee service for verification of actions generated by a set of AI agents related to a task that the set of AI agents are participating in, wherein the instructions, when executed by the at least one processor, cause the apparatus to perform operations, the operations comprising:receiving, from a user equipment, a request for verification of a first action, the request comprising information related to the first action, the first action being related to a task that a set of artificial intelligence (AI) agents are participating in, the set of AI agents including at least an AI agent on the user equipment;in response to receiving the request, performing a referee service to verify the first action; andsending, to the user equipment, a response to the request, wherein the response indicates a result of the verification.13.The apparatus of claim 12, wherein the operations further comprise:receiving, from the user equipment, information related to the set of AI agents, the information comprising an identity of the set of AI agents, a set of rules for the task, and identities of respective AI agents included in the set of AI agents; andsending, to the user equipment, a confirmation of receipt of the information related to the set of AI agents.14.The apparatus of claim 12 or 13, wherein the operations further comprise:receiving, from the user equipment, a request for establishment of the referee service for verification of actions related to the task generated by AI agents included in the set of AI agents;in response to receiving the request, establishing the referee service; andsending, to the user equipment, a response to the request, the response indicating the establishment of the referee service.15.The apparatus of claim 14, wherein establishing the referee service comprises:determining whether one or more user equipments associated with the set of AI agents have subscribed to the referee service; andin accordance with a determination that the one or more user equipments have subscribed to the referee service, establishing the referee service.16.The apparatus of claim 15, wherein establishing the referee service further comprises:assigning a global identity to the set of AI agents.17.The apparatus of claim 12, wherein the operations further comprise:creating the set of AI agents, based on receipt of a trigger that indicates the set of AI agents is to be created;sending, to the user equipment, a request for the AI agent to be included in the set of AI agents participating in the task; andreceiving, from the user equipment, a response to the request, the response indicating consent of the AI agent to be included in the set of AI agents participating in the task.18.The apparatus of claim 12, wherein the operations further comprise :receive, from the user equipment, a request for the AI agent to join the set of AI agents participating in the task; andin response to receiving the request, send, to the user equipment, a response to the request, the response indicating approval of the request.19.The apparatus of any of claims 12 to 18, wherein the operations further comprise:sending, to one or more user equipments associated with the set of AI agents, capability information of the network function, the capability information including information indicating that the network function is capable of establishment of the referee service and information about capabilities of the referee service.20.The apparatus of any of claims 12 to 19, wherein the information related to the first action comprises a context of the user equipment and the task associated with the first action.21.The apparatus of any of claims 12 to 20, wherein the operations further comprise :receiving, from the user equipment, a request for the referee service to be terminated;in response to receiving the request, terminating the referee service; andsending, to the user equipment, an indication that the referee service has been terminated.22.The apparatus of claim 21, wherein the request includes information indicative of a reason for the termination of the referee service.23.The apparatus of claims 12 to 22, wherein the verification of the first action is based on coordination between the network function and another network function of the communication network, wherein both the network functions are configured to provide the referee service for the verification of actions generated by the set of AI agents.24.The apparatus of claim 23, wherein the coordination between the network functions is controlled by a further network function.25.A method comprising:at a user equipment,obtaining, from an AI agent on the user equipment, a first action generated by the AI agent, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least the AI agent;sending, to a network function of a communication network, a request for verification of the first action, the request comprising information related to the first action, wherein the request causes a referee service of the network function to verify the first action; andreceiving a response to the request, wherein the response indicates a result of the verification.26.A method comprising:at a network function,receiving, from a user equipment, a request for verification of a first action, the request comprising information related to the first action, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least an AI agent on the user equipment;in response to receiving the request, performing a referee service to verify the first action; andsending, to the user equipment, a response to the request, wherein the response indicates a result of the verification.27.A first apparatus comprising:means for obtaining, from an AI agent on the user equipment, a first action generated by the AI agent, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least the AI agent;means for sending, to a network function of a communication network, a request for verification of the first action, the request comprising information related to the first action, wherein the request causes a referee service of the network function to verify the first action; andmeans for receiving a response to the request, wherein the response indicates a result of the verification.28.A second apparatus comprising:means for receiving, from a user equipment, a request for verification of a first action, the request comprising information related to the first action, the first action being related to a task that a set of AI agents are participating in, the set of AI agents including at least an AI agent on the user equipment;means for in response to receiving the request, performing a referee service to verify the first action; andmeans for sending, to the user equipment, a response to the request, wherein the response indicates a result of the verification.29.A computer readable medium comprising instructions stored thereon for causing an apparatus at least to perform the method of claim 25 or the method of claim 26.
