Collaboration method and apparatus between agents

Abstract

The application provides a method and device for cooperation between agents, and cooperation between multiple agents can be realized. The method comprises the following steps: acquiring first information, wherein the first information comprises information for indicating a first task; determining N agents for cooperating to complete the first task according to the first information, wherein N is a positive integer greater than 1; and sending first behavior information to a first agent, wherein the first agent is included in the N agents, and the first behavior information is used for indicating the behavior of the first agent when cooperating with other agents except the first agent in the N agents to complete the first task.

Description

Technical Field

[0001] This application relates to the field of communications, and more particularly to a method and apparatus for collaboration between intelligent agents. Background Technology

[0002] An intelligent agent is an entity with a certain level of intelligence and behavioral capabilities, capable of perceiving its environment, making decisions, and executing actions. However, a single intelligent agent often struggles to handle various complex tasks, thus leading to the concept of multi-agent collaboration. Multi-agent collaboration refers to multiple intelligent agents working together to complete a task.

[0003] However, no specific technical solution has yet been proposed for achieving collaboration among multiple cross-ecological intelligent agents in a wide-area environment. Summary of the Invention

[0004] This application provides a method and apparatus for collaboration among intelligent agents to enable multiple intelligent agents to collaboratively complete tasks.

[0005] Firstly, this application provides a method for cooperation between intelligent agents. This method can be executed, for example, by an agent communication network-control plane (ACN-C) network element, or by a component (such as a chip, chip system, etc.) configured in the ACN-C network element, or by a logic module or software capable of implementing all or part of the functions of the ACN-C network element. This application does not limit the scope of this method. In this application, the ACN-C network element is used as an example for description.

[0006] The method includes: acquiring first information, the first information including information for instructing a first task; determining N intelligent agents to collaboratively complete the first task based on the first information, where N is a positive integer greater than 1; sending first action information to the first intelligent agent, the first intelligent agent being included among the N intelligent agents, the first action information being used to instruct the behavior of the first intelligent agent when collaborating with other intelligent agents among the N intelligent agents to complete the first task.

[0007] In this application, the N intelligent agents that collaborate to complete the first task are also referred to as the N team members who collaborate to complete the first task.

[0008] Based on the technical solution provided in the first aspect, after receiving the first information, the ACN-C network element will determine the N intelligent agents that can cooperate to complete the first task and the corresponding behavioral information of each intelligent agent. Then, it will send the corresponding behavioral information to each of the N intelligent agents so that the N intelligent agents can cooperate to complete the first task.

[0009] In conjunction with the first aspect, in one possible implementation, the first behavioral information is used to instruct the behavior of the first agent when collaborating with other agents among N agents (excluding the first agent) to complete a first task, including:

[0010] The first line of information is used to indicate at least one of the following: the collaboration method when the first agent collaborates with other agents to complete the first task, the role undertaken by the first agent, the message template when the first agent communicates with other agents, the identifiers of other agents, and the capabilities of other agents.

[0011] For example, information in the first action information that can indicate the collaboration method when the first intelligent agent collaborates with other intelligent agents to complete the first task is also called collaboration method information; information in the first action information that can indicate the role the first intelligent agent determines is also called role information; information in the first action information that can indicate the message template when the first intelligent agent communicates with other intelligent agents is also called message template information; information in the first action information that can indicate the identifier of other intelligent agents is also called identifier information; and information in the first action information that can indicate the capabilities of other intelligent agents is also called capability information. That is, in one implementation, the first action information may include at least one of the following: collaboration method information, role information, message template information, identifier information, and capability information; the collaboration method information is used to indicate the collaboration method when the first intelligent agent collaborates with other intelligent agents to complete the first task; the role information is used for the first intelligent agent to determine the role it undertakes; the message template information is used to indicate the message template when the first intelligent agent communicates with other intelligent agents; the identifier information is used to indicate the identifier of other intelligent agents; and the capability information is used to indicate the capabilities of other intelligent agents.

[0012] In one possible implementation, the first behavioral information is used to indicate the collaboration method when the first intelligent agent collaborates with other intelligent agents to complete the first task, including: the first behavioral information indicates that the collaboration method when the first intelligent agent collaborates with other intelligent agents to complete the first task is a centralized collaboration method. For example, the first behavioral information includes collaboration method information, which indicates that the first intelligent agent adopts a centralized collaboration method when collaborating with other intelligent agents to complete the first task.

[0013] In this application, when the first action information indication adopts a centralized collaborative approach, the role undertaken by the first intelligent agent is as follows:

[0014] In one implementation, the ACN-C network element automatically determines the role of the first intelligent agent, and then the first action information includes instruction information indicating the role assumed by the first intelligent agent. For example, the first action information includes instruction information indicating that the first intelligent agent assumes the role of leader. Another example is that the first action information includes instruction information indicating that the first intelligent agent assumes the role of member.

[0015] In another implementation, the first action information includes a recommended leader instruction, which instructs the first agent to determine its role based on the recommended leader process. In other words, in this implementation, the ACN-C network element does not predetermine the role of the first agent but sends the recommended leader instruction to it, allowing the first agent to determine its own role based on the recommended leader process. Optionally, in this implementation, the first action information also includes a first winning rule. The first agent can determine its role based on the first winning rule. For example, the first agent can determine the agent assuming the leader role based on the first winning rule, thereby further determining its own role. For instance, if the first agent determines itself as the leader based on the recommended leader process and the first winning rule, then the first agent determines its own role as the leader; if the first agent determines another agent as the leader based on the recommended leader process and the first winning rule, then the first agent determines its own role as a member.

[0016] In conjunction with the first aspect, in one possible implementation, the first behavioral information is used to indicate the collaborative mode when the first intelligent agent collaborates with other intelligent agents to complete the first task, including: the first behavioral information is used to indicate that the collaborative mode when the first intelligent agent collaborates with other intelligent agents to complete the first task is a distributed collaborative mode.

[0017] When the first line of information indicates that a distributed collaborative approach is adopted, the first line of information is also used to indicate at least one of the following: the speaking order of the N agents, the arbitration method when at least two of the N agents have a decision conflict, the data mode used when the N agents negotiate, and the maximum token length used when the N agents negotiate.

[0018] Arbitration can also be described as a conflict resolution method or a conflict management method.

[0019] In this application, the information in the first action information that indicates the data mode used during negotiation among N agents and / or the maximum token length used during negotiation among N agents is also referred to as model negotiation capability information. Similarly, the information in the first action information that indicates the speaking order of N agents and / or the arbitration method when at least two of the N agents have a decision conflict is also referred to as dialogue flow control information. For example, in one implementation, the first action information may include at least one of the following: dialogue flow control information and model negotiation capability information; the dialogue flow control information is used to indicate the speaking order of N agents and / or the arbitration method when at least two of the N agents have a decision conflict, and the model negotiation capability information is used to indicate the data mode used during negotiation among N agents and / or the maximum token length used during negotiation among N agents.

[0020] The first line of information is used to indicate the speaking order of the N agents:

[0021] In one implementation, the first line of information is used to indicate that speaking should be done in a fixed order. Correspondingly, each agent speaks in a fixed order.

[0022] In another implementation, the first line of information is used to instruct speaking according to the speaking manager's schedule. Here, the speaking manager refers to the agent that manages the speaking of N agents. Correspondingly, each agent speaks when scheduled by the speaking manager. It is understandable that the speaking manager is one of the N agents.

[0023] The first line of information is used to indicate the arbitration method when at least two of the N agents have a decision conflict:

[0024] In one implementation, the first line of information is used to instruct arbitration based on the priorities of at least two agents. Correspondingly, when the decisions of at least two agents conflict, the conflict is managed according to the priorities of these at least two agents. For example, when the decisions of two agents conflict, the decision of the higher-priority agent is retained, and the decision of the lower-priority agent is re-executed until there is no conflict between the two agents.

[0025] In another implementation, the first line of information is used to instruct arbitration to be conducted based on the arbitration manager. The arbitration manager can also be referred to as a conflict manager, conflict resolver, etc.

[0026] In this application, when the first action information indication adopts a distributed cooperative approach, the role assumed by the first intelligent agent in indicating the first action information is as follows:

[0027] In one implementation, the ACN-C network element determines the role undertaken by the first intelligent agent, and then includes indication information in the first action information to indicate the role undertaken by the first intelligent agent.

[0028] For example, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a member.

[0029] For example, if the first action information is used to instruct the speaker to speak according to the scheduling of the speaker manager, and the ACN-C network element determines that the first intelligent agent assumes the role of the speaker manager, then the first action information includes instruction information for instructing the first intelligent agent to assume the role of the speaker manager.

[0030] For example, if the first action information is used to instruct arbitration based on the arbitration manager, and the ACN-C network element determines that the first intelligent agent assumes the role of the arbitration manager, the first action information includes instruction information for instructing the first intelligent agent to assume the role of the arbitration manager.

[0031] In another implementation, if the first action information is used to instruct speaking according to the scheduling of the speaking manager, the first action information includes a recommended speaking manager instruction, which instructs the first intelligent agent to determine its role based on the recommended speaking manager process; and / or, if the first action information is used to instruct arbitration based on the arbitration manager, the first action information includes a recommended arbitration manager instruction, which instructs the first intelligent agent to determine its role based on the recommended arbitration manager process. In other words, in this implementation, the ACN-C network element sends recommended speaking manager instructions and / or recommended arbitration manager instructions to the first intelligent agent, enabling the first intelligent agent to determine its role based on the recommended leader process.

[0032] For example, when the first action information includes instructions to recommend a speaking manager, it also includes a second winning rule. The first agent can determine its own role based on this second winning rule. For instance, the first agent can determine which agent will assume the role of a speaking manager based on the second winning rule, thereby further determining its own role.

[0033] For example, when the first action information includes instructions to recommend an arbitration manager, it also includes a third winning rule. The first agent can determine its own role based on this third winning rule. For instance, the first agent can determine the agent assuming the role of recommending an arbitration manager based on the third winning rule, thereby further determining its own role.

[0034] In conjunction with the first aspect, in one possible implementation, the method further includes: receiving first capability information from a first intelligent agent, the first capability information being used to indicate the capabilities of the first intelligent agent, the capabilities including ontological capabilities and / or model capabilities.

[0035] In conjunction with the first aspect, in one possible implementation, obtaining the first information includes: receiving the first information from the first intelligent agent.

[0036] In conjunction with the first aspect, in one possible implementation, the first information further includes one or more of the following: identifiers of candidate intelligent agents that the first intelligent agent expects to cooperate in completing the first task; and indication information for indicating the cooperative method expected by the first intelligent agent.

[0037] It should be noted that this application does not restrict whether the ACN-C network element determines N agents based on the candidate agents expected by the first agent after receiving the identifiers of the candidate agents expected by the first agent. For example, after the ACN-C network element receives the identifiers of the candidate agents expected by the first agent, in one implementation, the ACN-C network element selects the aforementioned N agents from the candidate agents expected by the first agent; in another implementation, the ACN-C network element does not select the aforementioned N agents from the candidate agents expected by the first agent, that is, the ACN-C determines the N agents to cooperate in completing the first task independently.

[0038] In conjunction with the first aspect, in one possible implementation, the first intelligent agent expects a centralized collaborative approach, and the first information also includes instruction information indicating that the first intelligent agent expects to assume a leader role.

[0039] Secondly, this application provides a method for collaboration between intelligent agents. This method can be executed by a first intelligent agent, or by a component (such as a chip, chip system, etc.) configured in the first intelligent agent, or by a logic module or software capable of implementing all or part of the functions of the first intelligent agent. This application does not limit the scope of this method. In this application, the first intelligent agent is used as an example for description.

[0040] The method includes: receiving first action information, the first action information being used to instruct the behavior of a first agent when collaborating with other agents among N agents (excluding the first agent) to complete the first task, where N is a positive integer greater than 1; and collaborating with other agents among the N agents to complete the first task according to the first action information.

[0041] In conjunction with the second aspect, in one possible implementation, the first behavioral information is used to instruct the behavior of the first agent when collaborating with other agents among N agents to complete the first task, including: the first behavioral information is used to instruct at least one of the following: the collaboration method when the first agent collaborates with other agents to complete the first task, the role undertaken by the first agent, the message template when the first agent communicates with other agents, the identifiers of other agents, and the capabilities of other agents.

[0042] In conjunction with the second aspect, in one possible implementation, the first behavioral information is used to indicate the collaborative mode when the first intelligent agent collaborates with other intelligent agents to complete the first task, including: the first behavioral information is used to indicate that the collaborative mode when the first intelligent agent collaborates with other intelligent agents to complete the first task is a centralized collaborative mode.

[0043] In conjunction with the second aspect, in one possible implementation, the first behavioral information is used to indicate the role assumed by the first intelligent agent, including: the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a leader; or, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a member; or, the first behavioral information includes leader recommendation instruction information, which is used to instruct the first intelligent agent to determine the role assumed by the first intelligent agent based on the leader recommendation process.

[0044] In conjunction with the second aspect, in one possible implementation, if the first action information includes recommended leader instruction information, the first action information also includes a first winning rule; wherein, the first agent determines the role it assumes based on the first winning rule.

[0045] In conjunction with the second aspect, in one possible implementation, the first behavioral information is used to indicate the collaborative mode when the first intelligent agent collaborates with other intelligent agents to complete the first task, including: the first behavioral information is used to indicate that the collaborative mode when the first intelligent agent collaborates with other intelligent agents to complete the first task is a distributed collaborative mode.

[0046] In conjunction with the second aspect, in one possible implementation, the first action information is also used to indicate at least one of the following: the speaking order of the N agents, the arbitration method when at least two of the N agents have a decision conflict, the data mode used when the N agents negotiate, and the maximum token length used when the N agents negotiate.

[0047] In conjunction with the second aspect, in one possible implementation, the first action information is used to indicate the speaking order of the N agents, including: the first action information is used to indicate speaking in a fixed speaking order; or, the first action information is used to indicate speaking according to the scheduling of the speaking manager.

[0048] In conjunction with the second aspect, in one possible implementation, the first behavioral information is used to indicate the arbitration method when at least two of the N agents have a decision conflict, including: the first behavioral information is used to indicate arbitration based on the priority of the at least two agents; or, the first behavioral information is used to indicate arbitration based on the arbitration manager.

[0049] In conjunction with the second aspect, in one possible implementation, the first behavioral information is used to instruct the first intelligent agent to determine the role it assumes, including: the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a member; and / or, if the first behavioral information is used to instruct speaking according to the scheduling of the speaking manager, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a speaking manager; and / or, if the first behavioral information is used to instruct arbitration based on the arbitration manager, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of an arbitration manager.

[0050] In conjunction with the second aspect, in one possible implementation, the first action information is used to instruct the first intelligent agent to determine the role it assumes, including: if the first action information is used to instruct speaking according to the scheduling of the speaking manager, the first action information includes a recommended speaking manager instruction, which instructs the first intelligent agent to determine the role it assumes based on the process of recommending the speaking manager; and / or, if the first action information instructs arbitration based on the arbitration manager, the first action information includes a recommended arbitration manager instruction, which instructs the first intelligent agent to determine the role it assumes based on the process of recommending the arbitration manager.

[0051] In conjunction with the second aspect, in one possible implementation, if the first action information is used to instruct speaking according to the scheduling of the speaking manager, the step of cooperating with other intelligent agents among the N intelligent agents to complete the first task includes: receiving speaking instruction information from the speaking manager, the speaking instruction information being used to instruct the first intelligent agent to speak.

[0052] In conjunction with the second aspect, in one possible implementation, cooperating with other agents among the N agents to complete the first task includes: sending a first conflict indication message, which indicates that the decision result of the first agent conflicts with the decision result of a second agent among the N agents; and / or receiving a second conflict indication message, which indicates that the decision result of the first agent conflicts with the decision result of a third agent among the N agents; and / or receiving a re-decision indication message, which instructs the first agent to re-make a decision.

[0053] Thirdly, this application provides a collaborative device that has the functions of the first aspect described above. For example, the device includes modules, units, or means that perform the operations involved in the first aspect. These modules, units, or means can be implemented by software, hardware, or a combination of software and hardware.

[0054] The device is, for example, an ACN-C network element, or a device that includes ACN-network element functionality, or a chip system (or chip or circuit) or other functional module that can implement the ACN-network element functionality, and the chip system or functional module is, for example, located in the ACN-network element.

[0055] In one possible implementation, the device includes a processing unit (sometimes also called a processing module) and a transceiver unit (sometimes also called a transceiver module). The transceiver unit is capable of both sending and receiving functions. When the transceiver unit performs the sending function, it can be called a sending unit (sometimes also called a sending module); when the transceiver unit performs the receiving function, it can be called a receiving unit (sometimes also called a receiving module). The sending unit and the receiving unit can be the same functional module, which is called the transceiver unit and can perform both sending and receiving functions; alternatively, the sending unit and the receiving unit can be different functional modules, and the transceiver unit is a collective term for these functional modules.

[0056] In one possible implementation, the transceiver module is configured to: acquire first information, the first information including information for instructing a first task; the processing module is configured to: determine N intelligent agents that will collaboratively complete the first task based on the first information, where N is a positive integer greater than 1; the transceiver module is further configured to: send first behavior information to the first intelligent agent, the first intelligent agent being included among the N intelligent agents, the first behavior information being used to instruct the behavior of the first intelligent agent when collaborating with other intelligent agents among the N intelligent agents to complete the first task.

[0057] Fourthly, this application provides a collaborative device that has the functions of the second aspect above. For example, the device includes modules, units, or means that perform the operations involved in the second aspect above. These modules, units, or means can be implemented by software, hardware, or a combination of software and hardware.

[0058] The communication device is, for example, a first intelligent agent, or other device including the functions of a first intelligent agent, or a chip system (or, chip or circuit) or other functional module, which is capable of realizing the functions of the first intelligent agent, and the chip system or functional module is, for example, set in the first intelligent agent.

[0059] In one possible implementation, the device includes a processing unit (sometimes also called a processing module) and a transceiver unit (sometimes also called a transceiver module). The transceiver unit is capable of both sending and receiving functions. When the transceiver unit performs the sending function, it can be called a sending unit (sometimes also called a sending module); when the transceiver unit performs the receiving function, it can be called a receiving unit (sometimes also called a receiving module). The sending unit and the receiving unit can be the same functional module, which is called the transceiver unit and can perform both sending and receiving functions; alternatively, the sending unit and the receiving unit can be different functional modules, and the transceiver unit is a collective term for these functional modules.

[0060] In one possible implementation, the transceiver module is configured to: receive first behavioral information from an ACN-C network element, the first behavioral information being used to instruct the behavior of the first agent when collaborating with other agents among N agents (excluding the first agent) to complete a first task, where N is a positive integer greater than 1; and the processing module is configured to: collaborate with other agents among the N agents to complete the first task based on the first behavioral information.

[0061] Fifthly, this application provides an apparatus including a processor for performing the methods described in any of the foregoing aspects and any possible implementations of any of the foregoing aspects.

[0062] In one possible implementation, the processor is configured to, by executing a computer program or instructions, and / or, by logic circuitry, cause the communication device to perform the method described in any of the above aspects and any possible implementations of any of the aspects.

[0063] The apparatus may further include a memory for storing instructions and / or data. The memory is coupled to the processor, which, when executing the instructions stored in the memory, can implement the methods described in the foregoing aspects. Optionally, the memory and the processor are integrated together.

[0064] The device may also include a communication interface for communicating with other devices. For example, the communication interface may be a transceiver, circuit, bus, module or other type of communication interface.

[0065] Sixthly, this application provides a chip system including at least one processor for supporting the functions involved in any of the above aspects and any possible implementations of any aspect.

[0066] In one possible design, the chip system also includes a memory for storing program instructions and data, which may be located within or outside the processor.

[0067] The chip system can consist of chips or include chips and other discrete components.

[0068] In a seventh aspect, this application provides a computer-readable storage medium including a computer program that, when run on a computer, causes the computer to implement the methods in any of the foregoing aspects and any possible implementations of any of the foregoing aspects.

[0069] Eighthly, this application provides a computer program product comprising: a computer program (also referred to as code or instructions) that, when run, causes a computer to perform the methods in any of the foregoing aspects and any possible implementations of any of the foregoing aspects.

[0070] Ninthly, this application provides a communication system including the aforementioned ACN-C network element and a first intelligent agent. The ACN-C network element is used to instruct the methods in the first aspect and any possible implementation thereof; the first intelligent agent is used to execute the methods in the second aspect and any possible implementation thereof. Attached Figure Description

[0071] Figure 1 A schematic diagram of an intelligent agent communication network is shown.

[0072] Figure 2 The system architecture to which this application applies is illustrated;

[0073] Figure 3 A flowchart illustrating a collaborative method provided in one embodiment of this application;

[0074] Figures 4-6 A flowchart illustrating the collaborative process using a centralized approach provided in this application;

[0075] Figures 7-8 A flowchart illustrating the collaborative process using a distributed collaborative approach provided in this application;

[0076] Figure 9 A flowchart illustrating the process of speaking in a fixed order when using a distributed collaborative approach, as provided in this application;

[0077] Figure 10 A flowchart illustrating the process of scheduling speeches based on a speech manager when using a distributed collaborative approach, as provided in this application;

[0078] Figure 11 A flowchart illustrating the process of conflict management based on agent priority when using a distributed collaborative approach, as provided in this application.

[0079] Figure 12A flowchart illustrating the process of conflict management based on arbitration managers when using a distributed collaborative approach, as provided in this application;

[0080] Figure 13 A schematic diagram illustrating the handling process for speech anomalies during collaborative work using a distributed collaborative approach, as provided in this application;

[0081] Figure 14 This is a flowchart illustrating a collaborative device provided in one embodiment of this application;

[0082] Figure 15 This is a flowchart illustrating a collaborative device provided in another embodiment of this application. Detailed Implementation

[0083] First, let me introduce some of the terms used in this application.

[0084] 1. Intelligent agent

[0085] An intelligent agent is an entity with a certain level of intelligence and behavioral capabilities, possessing basic characteristics such as autonomy, interactivity, responsiveness, and adaptability. It is able to perceive its environment, make decisions, and execute actions.

[0086] Autonomy: Intelligent agents are able to make decisions and act autonomously without human intervention. For example, in a smart home system, an intelligent agent can automatically adjust the air conditioner and humidifier based on indoor temperature and humidity to provide the most comfortable living environment.

[0087] Interactivity: Intelligent agents are able to communicate and cooperate with other intelligent agents or humans. For example, in autonomous vehicles, intelligent agents need to interact with traffic lights, other vehicles, and pedestrians to ensure safe driving.

[0088] Responsiveness: An intelligent agent can respond promptly to external stimuli. For example, in the field of robotics, an intelligent agent needs to perceive changes in the environment and take corresponding actions to adapt.

[0089] Adaptability: Intelligent agents can adjust their behavior based on experience to adapt to complex and ever-changing environments.

[0090] Intelligent agents have a wide range of applications, including automated systems, robots, virtual assistants, and game characters. They can complete various tasks independently or collaborate with other intelligent agents or humans to accomplish complex tasks. For example, intelligent agents can be mobile phones, tablets, computers with wireless transceiver capabilities, wearable devices, vehicles, airplanes, ships, robots, robotic arms, smart home devices, and more.

[0091] 2. Agent Network

[0092] An agent communication network (ACN) refers to a communication network composed of multiple intelligent agents. An ACN can provide a secure and controllable communication network for various intelligent agents, offering on-demand, secure, and controllable information interaction services. For example, such as... Figure 1 As shown, this includes multiple intelligent agents, such as the intelligent agents in home / construction site #1, the intelligent agents in home / construction site #2, the digital life assistant intelligent agent, and the intelligent agents in the smart construction site / factory. Each intelligent agent can access the ACN subnet to communicate with other intelligent agents. For example, as... Figure 1 As shown, the robotic vacuum cleaner in Home / Construction Site #1 can be connected to subnet 1 in ACN, the robot dog in Home / Construction Site #1 can be connected to subnets 1 and 2 in ACN, and the drone in Home / Construction Site #1 can be connected to subnets 1 and 3 in ACN.

[0093] In complex information environments, a single agent often struggles to handle various complex tasks. Therefore, multi-agent collaboration has been proposed. Multi-agent collaboration refers to the process by which multiple agents work together to complete tasks or achieve goals through cooperation, coordination, and communication. The importance and advantages of this collaborative mechanism are mainly reflected in the following aspects: 1) Improved efficiency: Multi-agent collaborative work can significantly improve the speed and efficiency of task execution, especially in large-scale data processing and analysis scenarios. 2) Enhanced robustness: While a single agent may fail for various reasons, the redundant design of a multi-agent system can ensure the stable operation of the overall system. 3) Strong adaptability: Faced with constantly changing needs and environments, multi-agent systems can quickly adjust their strategies and flexibly respond to various situations.

[0094] However, no specific technical solution has yet been proposed for how to achieve multi-agent collaboration.

[0095] In view of this, this application provides a method and apparatus for collaboration among intelligent agents to achieve collaboration among multiple intelligent agents. The technical solution provided in this application is described below.

[0096] refer to Figure 2 , Figure 2 The system architecture to which the technical solution of this application applies is illustrated. For example... Figure 2 As shown, the architecture includes an agent, an access network (AN), an agent communication network-user plane (ACN-U) network element, and an agent communication network-control plane (ACN-C) network element.

[0097] The concept of an intelligent agent can be found in the description above, and will not be repeated here.

[0098] This application does not limit the specific implementation form of the intelligent agent. For example, it can be an intelligent agent as a terminal, such as a robot or intelligent vehicle. A terminal can also be called an access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile equipment, user terminal, wireless telecom equipment, user agent, user equipment, or user device. Another example is an intelligent agent deployed in a carrier network as a network-based intelligent service, such as a personal voice assistant. Yet another example is an intelligent agent deployed in the cloud.

[0099] In this application, intelligent agents can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), Internet of Things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearables, smart transportation, smart cities, etc.

[0100] The primary function of the AN is to control the intelligent agent to wirelessly access the ACN. For example, the AN can be any device with wireless transceiver capabilities. This equipment includes, but is not limited to: base stations, evolved NodeBs (eNBs or eNodeBs), radio network controllers (RNCs), Node Bs (NBs), base station controllers (BSCs), base transceiver stations (BTSs), home base stations (e.g., home-evolved NodeBs or home Node Bs, HNBs), base band units (BBUs), access points (APs), wireless relay nodes, wireless backhaul nodes, transmission points (TPs), or transmission and reception points (TRPs) in Wi-Fi systems. It can also be a gNB in ​​5G systems, such as NR, or a transmission point (TRP or TP), one or a group of antenna panels (including multiple antenna panels) of a base station in a 5G system, or a network node constituting a gNB or transmission point, such as a base band unit (BBU) or a distributed unit (DMU). Unit (DU), etc.; or, it can be the next-generation base station in the 6th generation (6G) mobile communication system, the base station in the future mobile communication system, or the access node in the WiFi system, etc.

[0101] The ACN-U network element is the user plane network element of the ACN, responsible for user plane data exchange between agents and deploying agents in the network as needed to participate in decision negotiation. The ACN-C network element is the control plane network element of the ACN, responsible for agent access authentication, session management, and authorization for decision negotiation mechanisms.

[0102] It should be noted that the names of the network elements involved in the embodiments of this application can be either the names of the network elements mentioned in the above embodiments or the names of network elements with the same functions as the above network elements in a future communication system, and do not constitute a limitation of this application.

[0103] The technical solution provided in this application will now be described with reference to the accompanying drawings.

[0104] refer to Figure 3 , Figure 3This is a flowchart illustrating a communication method provided in one embodiment of this application. Figure 3 As shown, the method includes:

[0105] S310, the ACN-C network element obtains first information, which includes information used to instruct the first task.

[0106] In this application, the information used to indicate the first task in the first information is also referred to as task description information.

[0107] In one implementation, the ACN-C network element acquiring the first information includes: the ACN-C network element receiving the first information sent by the intelligent agent. In this application, the intelligent agent sending the first information is also referred to as the requesting intelligent agent.

[0108] Optionally, the first information may also include the identifiers of candidate intelligent agents that the first intelligent agent expects (or may request) to cooperate in completing the first task. That is, the first information can also be understood as indicating the candidate intelligent agents that the first intelligent agent expects to cooperate in completing the first task.

[0109] Optionally, the first information may also include information indicating the desired collaboration method of the first intelligent agent, also known as collaboration method request information. In other words, the first information is also used to indicate the desired collaboration method of the first intelligent agent. In this application, the collaboration method may be a centralized collaboration method or a distributed collaboration method. Optionally, when the first information indicates that the desired collaboration method is a centralized collaboration method, the first information may also include indication information indicating that the first intelligent agent expects to assume a leader role.

[0110] In another implementation, the ACN-C network element acquiring the first information includes: the ACN-C network receiving the first information sent by the user terminal (such as a mobile phone).

[0111] Optionally, the first information may also include the identifiers of candidate intelligent agents that the user expects to collaborate in completing the first task. That is, the first information can also be understood as indicating candidate intelligent agents that the user expects to collaborate in completing the first task.

[0112] Optionally, the first information also includes information indicating the user's desired collaboration method, also known as collaboration method request information. That is, the first information is also used to indicate the user's desired collaboration method. In this application, the collaboration method can be a centralized collaboration method or a distributed collaboration method. Optionally, when the first information indicates that the desired collaboration method is a centralized collaboration method, the first information also includes indication information indicating that the candidate intelligent agent is expected to assume the leader role.

[0113] S320, the ACN-C network element determines N intelligent agents to collaboratively complete the first task, where N is a positive integer greater than 1.

[0114] In this application, the N intelligent agents that are determined to cooperate in completing the first task are also referred to as the N team members.

[0115] The ACN-C network element can determine N intelligent agents that collaboratively complete the first task in different ways.

[0116] For example, in one implementation, the ACN-C network element determines the N agents required to complete the first task based on the capabilities of each agent registered with the ACN-C network element. The capabilities of the agents include their ontological capabilities and / or their model capabilities.

[0117] For example, the agent's capabilities may include one or more of the following: agent type, operational capabilities, movement capabilities, and perception capabilities. This application does not limit the specific content of operational and movement capabilities. For example, operational capabilities may include grasping multiple objects or operational precision; perception capabilities may include lidar detection capabilities or camera perception capabilities. For example, movement capabilities may include driving speed, two-wheeled movement, or four-wheeled movement.

[0118] For example, an agent's model capabilities include one or more of the following: whether it supports large language models (LLMs), the type of LLM, the longest token length supported by the LLM, and / or the data modality type. The type of LLM can be one or more of the following: autoregressive language model, autoencoder language model, encoder-decoder structure model. It is important to emphasize that this is only a partial list of LLM types; more types of LLMs may emerge as technology advances. The meaning of data modality type is as follows: Data is the foundation of artificial intelligence (AI). Data sources are wide-ranging and diverse in form; each source or form can be considered a modality, such as video, images, text, speech, and sensor data and sound spectra in industrial scenarios.

[0119] Optionally, if the first information also indicates candidate agents for collaboratively completing the first task, the ACN-C network element can also select N agents from the expected candidate agents to complete the first task. However, it should be noted that when the first information indicates candidate agents, this application does not restrict whether the ACN-C network element specifically selects the N agents needed to complete the first task from the candidate agents. For example, in another implementation, although the first information indicates candidate agents for collaboratively completing the first task, the ACN-C network element does not select the aforementioned N agents from the candidate agents; that is, the ACN-C network element determines the aforementioned N agents itself.

[0120] Optionally, the ACN-C network element can also create network agents within the ACN-U network element, and include the network agents created in the ACN-U network element among the aforementioned N agents. That is, the N agents that the ACN-C network element determines will collaboratively complete the first task may include the network agents created by the ACN-C network element within the ACN-U network element. In one implementation, the network agents created in the ACN-U network element can be used for message monitoring / management among team members.

[0121] S330, the ACN-C network element sends first action information to the first intelligent agent. The first intelligent agent is included among N intelligent agents. The first action information is used to instruct the behavior of the first intelligent agent when it collaborates with other intelligent agents among the N intelligent agents to complete the first task.

[0122] In this application, after identifying the N intelligent agents that will collaboratively complete the first task, the ACN-C network element determines the behavioral information corresponding to each of the N intelligent agents and then sends the corresponding behavioral information to each intelligent agent. In this application, the behavioral information can also be replaced by pattern information, behavioral pattern information, collaborative information, etc.

[0123] Specifically, for the first intelligent agent among N intelligent agents, the first behavioral information is used to instruct the first intelligent agent to cooperate with the other intelligent agents among the N intelligent agents to complete the first task. The behavior can also be replaced by a mode, coordination mode, etc. Correspondingly, the first intelligent agent cooperates with the other intelligent agents among the N intelligent agents to complete the first task based on the instructions of the first behavioral information.

[0124] In one example, the first agent is the requesting agent. That is, the first agent is also the agent that sends the first information to the ACN-C network element.

[0125] In one implementation, the first behavioral information is used to instruct the behavior of the first agent when collaborating with other agents among N agents to complete a first task, including: the first behavioral information instructs at least one of the following: the collaborative method when the first agent collaborates with other agents to complete the first task; the role undertaken by the first agent; the message template when the first agent communicates with other agents; the identifiers of other agents; and the capabilities of other agents.

[0126] In this application, the information in the first action information that can instruct the collaboration method when the first intelligent agent collaborates with other intelligent agents to complete the first task is also called collaboration method information; the information in the first action information that can instruct the first intelligent agent to determine its role is also called role information; the information in the first action information that can instruct the first intelligent agent to communicate with other intelligent agents is also called message template information; the information in the first action information that can instruct the identification of other intelligent agents is also called identification information; and the information in the first action information that can instruct the capabilities of other intelligent agents is also called capability information. For example, in one implementation, the first action information includes at least one of the following: identification information, capability information, message template information, collaboration method information, and role information. Wherein, identification information is used to indicate the identification of other intelligent agents that collaborate with the first intelligent agent to complete the first task. Capability information is used to indicate the capabilities of the aforementioned other intelligent agents. Message template information is used to indicate the message template when the first intelligent agent communicates with the aforementioned other intelligent agents. Collaboration method information is used to indicate the collaboration method when the first intelligent agent collaborates with the aforementioned other intelligent agents to complete the first task. Role information is used by the first intelligent agent to determine its role.

[0127] Correspondingly, after receiving the first action information, the first intelligent agent can learn about the capabilities of other intelligent agents and each intelligent agent in the other intelligent agents based on the identification information and capability information, learn about the collaboration method when collaborating with other intelligent agents to complete the first task based on the collaboration method information, learn about the message template used when communicating with other intelligent agents (the prompt that guides the LLM to generate semi-structured messages) based on the message template information, and learn about the role it plays when completing the first task based on the role information, that is, how it participates in the first task.

[0128] In this application, when the ACN-C network element indicates the coordination method to the first intelligent agent through the first line information, in one embodiment, the ACN-C network element indicates that a centralized coordination method is adopted.

[0129] For example, in one implementation, ACN-C determines to adopt a centralized collaboration approach when the following conditions are met: an agent within the team possesses the capabilities required to assume the Leader role in the first task, and the agent's contract also allows it to assume the Leader role. Here, the Leader role refers to the role capable of executing decisions.

[0130] Understandably, for N agents working together in a centralized collaborative manner to complete the first task, each agent can assume the role of either a leader or a member. The member role can be considered as following the leader's instructions.

[0131] In this application, when the ACN-C network element determines that the first intelligent agent and other intelligent agents adopt a centralized collaborative approach, and the ACN-C network element instructs the first intelligent agent to determine its role through first action information, there are two possible implementation schemes.

[0132] The first implementation scheme: The ACN-C network element determines the role of the first intelligent agent, and then includes indication information in the first action information to indicate the role (leader or member) undertaken by the first intelligent agent.

[0133] For example, if the capabilities of the first agent meet the requirements for assuming the Leader role in the first task, and the first agent's contract also allows it to assume the Leader role, then the ACN-C network element determines that the first agent assumes the Leader role; otherwise, the ACN-C network element determines that the first agent assumes the member role.

[0134] For example, the first agent is the requesting agent, and the first message sent by the first agent also includes information indicating that the first agent expects to assume the Leader role. In this case, if the ACN-C network element determines that the first agent's capabilities meet the requirements for assuming the Leader role in the first task, and the first agent's contract also allows it to assume the Leader role, then the ACN-C network element determines that the first agent assumes the Leader role.

[0135] In other words, under this implementation, the ACN-C network element determines the role to be undertaken by the first intelligent agent and instructs it.

[0136] The second implementation scheme: The first action information includes recommended leader instruction information, which is used to instruct the first intelligent agent to determine its role based on the recommended leader's process. In other words, the ACN-C network element does not determine the role of the first intelligent agent, but instead instructs the first intelligent agent to execute the recommended leader's process independently through the recommended leader instruction information, and then determines its role based on the recommended leader's process.

[0137] For example, in one implementation, the first action information may also include a first winning principle, which is used by the first agent to determine the Leader role among N agents. For example, the first winning principle includes: the agent with the most votes assumes the Leader role among the N agents; when the N agents have the same number of votes, the first speaker (the agent that recommended first) becomes the Leader among the N agents.

[0138] Understandably, in this implementation, for the first agent, if the first agent determines that it is the leader based on the leader recommendation process and the first winning principle information, then the first agent determines that it will assume the role of leader. If the first agent determines that one of the other N agents is the leader based on the leader recommendation process and the first winning principle information, then the first agent determines that it will assume the role of member.

[0139] In this application, when the ACN-C network element indicates the cooperative mode to the first intelligent agent through the first line information, in another embodiment, the ACN-C network element indicates that a distributed cooperative mode is adopted.

[0140] For example, in one implementation, the ACN-C network element determines to adopt a distributed collaborative approach when the following conditions are met: each of the N agents that collaboratively complete the first task has the ability to make distributed decisions and the contract also allows them to adopt a distributed collaborative approach.

[0141] Understandably, for an agent that uses a distributed collaborative approach to complete the first task, the roles it assumes include member roles, and in a distributed collaborative approach, each member can execute decisions.

[0142] In this application, when the ACN-C network element determines that the collaboration mode between the first agent and other team members is a distributed collaboration mode, the first action information is also used to indicate at least one of the following: the speaking order of the N agents, the arbitration mode when at least two of the N agents have a decision conflict, the data mode used when the N agents negotiate, and the maximum token length used when the N agents negotiate. The arbitration mode can also be described as a conflict resolution mode or a conflict management mode. In this application, the indication information in the first action information that can indicate the data mode used when the N agents negotiate and / or the maximum token length used when the N agents negotiate is also called model negotiation capability information, and the indication information in the first action information that can indicate the speaking order of the N agents and / or the arbitration mode when at least two of the N agents have a decision conflict is also called dialogue flow control information. For example, when the first action information indicates the use of a distributed collaboration mode, the first action information may also include at least one of the following: dialogue flow control information and model negotiation capability information. The model negotiation capability information is used to indicate the data mode used when the N agents negotiate and / or the maximum token length used when the N agents negotiate. Data modalities are typically the intersection of data modal types supported by at least two agents collaborating to complete the first task. The maximum length of supported tokens is typically determined by the agent with the smallest LLM-supporting token length among the N agents collaborating to complete the first task. Dialogue flow control information is used to indicate the speaking order of the N agents and / or the arbitration method when at least two of the N agents have decision conflicts. The speaking order can be considered as the order in which the N agents collaborating to complete the first task speak during the negotiation and discussion process to avoid speaking conflicts. Optionally, dialogue flow control information can also be used to indicate the negotiation depth, i.e., the maximum number of negotiations to complete the first task.

[0143] In this way, after receiving the first action information, the first agent will collaborate with other team members to complete the first task based on the indicated data modality, the negotiated maximum token length, the speaking order, and the arbitration method.

[0144] Specifically, the first line of information can be implemented in different ways to indicate the speaking order of N agents.

[0145] Implementation Method 1): The first line of information instructs N agents to speak in a fixed speaking order. Correspondingly, the N agents speak in a fixed speaking order. For example, the first line of information includes a speaking order configuration file, and each agent speaks according to the fixed order specified in that configuration file.

[0146] Implementation Method 2): The first line of information indicates that speaking should be done according to the scheduling of the chart manager.

[0147] For example, if the first action information sent to the first agent is to instruct it to speak according to the scheduling of the speech manager, then the first agent's speaking includes: speaking after receiving the speaking instruction information from the speech manager, whereby the speaking instruction information instructs the first agent to speak. Understandably, in this implementation, one of the N agents assumes the role of the speech manager. That is, the agent assuming the role of the speech manager can manage the speaking order of the agents collaboratively completing the first task. For example, if the N agents include an ACN-U network element, the ACN-U network element assumes the role of the speech manager.

[0148] Specifically, the first action information can be implemented in different ways when instructing the arbitration method when at least two agents have a decision conflict.

[0149] Implementation method 1): The first action information indicates arbitration based on the priorities of at least two agents.

[0150] For example, the first line of information includes the priorities of each of the N agents. When a conflict occurs between at least two of the N agents, the policy of the agent with the higher priority takes precedence.

[0151] Implementation Method 2): The first line of information indicates that arbitration should be conducted according to the arbitrator.

[0152] The first action information indicates that arbitration should be conducted by an arbitrator, also known as the management of decision conflicts by the arbitrator. Understandably, in this implementation, one of the N agents acts as the arbitrator. That is, the agent assuming the role of arbitrator can manage conflicts between agents collaborating to complete the first task. For example, if the N agents include a network agent created by an ACN-U network element, the network agent acts as the arbitrator. For instance, if the first action information indicates arbitration by the arbitrator, the arbitrator can send a re-decision instruction to the first agent. This re-decision instruction instructs the first agent to re-decide to resolve the conflict between the first agent and other agents.

[0153] In this application, if the first intelligent agent determines that the decisions of the second intelligent agent conflict, the first intelligent agent can send a first conflict indication message to the conflicting intelligent agent through an ACN-U network element. The first conflict indication message is used to indicate that the decision result of the first intelligent agent conflicts with the decision result of the second intelligent agent. In this application, the first intelligent agent can receive a second conflict indication message, which is used to indicate that the decision result of the first intelligent agent conflicts with the decision result of a third intelligent agent among N intelligent agents.

[0154] In this application, when the ACN-C network element determines that the first intelligent agent and other intelligent agents are working together in a distributed cooperative manner to complete the first task, the first action information indicating the role undertaken by the first intelligent agent can be implemented in different ways.

[0155] The first implementation scheme is as follows: The ACN-C network element determines the role of the first intelligent agent, and then includes instruction information in the first action information to indicate the role undertaken by the first intelligent agent.

[0156] In other words, under this implementation, the ACN-C network element determines the role to be undertaken by the first intelligent agent and instructs it.

[0157] Specifically, in this application, the first behavioral information includes instruction information for instructing the first intelligent agent to assume a member role.

[0158] Optionally, if the first action information indicates that speaking should be done according to the scheduling of the speaking manager, and if the ACN-C network element determines that the first intelligent agent is the speaking manager, then the first action information also includes instruction information for instructing the first intelligent agent to assume the role of the speaking manager.

[0159] Optionally, if the first action information indicates that arbitration should be conducted in accordance with the arbitration manager, and if the ACN-C network element determines that the first intelligent agent is the arbitration manager, then the first action information may also include instruction information for instructing the first intelligent agent to assume the role of arbitration manager.

[0160] The second implementation scheme: If the first line of information indicates that speaking is to be done according to the scheduling of the speaking manager, and / or arbitration is to be done according to the arbitration manager, the ACN-C network element does not determine the intelligent agent that will assume the role of speaking manager and / or arbitration manager. Instead, N intelligent agents will determine whether to assume the roles of speaking manager and arbitration manager, thereby determining their respective roles.

[0161] For example, if the first action information indicates speaking according to the speaking manager's schedule, then the first action information may include a recommendation of a speaking manager instruction. This instruction instructs the first agent to determine whether to assume the role of speaking manager based on the process of recommending a speaking manager. In one implementation, the first action information may also include a second winning principle, which is used by the first agent to determine the speaking manager role. For example, the second winning principle includes: the agent with the most votes is the speaking manager among the N agents; when the N agents have the same number of votes, the first speaker (the agent recommended first) is the speaking manager. Understandably, in this implementation, for the first agent, if the first agent determines that it is the speaking manager based on the process of recommending a speaking manager and the second winning principle information, then the first agent determines that its role includes the speaking manager role; if the first agent determines that another agent among the N agents is the speaking manager based on the process of recommending a speaking manager and the second winning principle information, then the first agent determines that its role does not include the speaking manager role.

[0162] For example, if the first action information indicates that arbitration should be conducted according to the arbitration manager, then the first action information may include an instruction to recommend an arbitration manager. This instruction instructs the first agent to determine whether it should assume the role of arbitration manager based on the process of recommending an arbitration manager. In one implementation, the first action information may also include a third winning principle, which is used by the first agent to determine the arbitration manager. For example, the third winning principle includes: the agent with the most votes is the arbitration manager; when N agents have the same number of votes, the first speaker (the agent who recommended the arbitration manager first) is the arbitration manager. Understandably, in this implementation, for the first agent, if the first agent determines that it is the arbitration manager based on the process of recommending an arbitration manager and the third winning principle information, then the first agent determines that its role includes the role of arbitration manager; if the first agent determines that another agent is the arbitration manager based on the process of recommending an arbitration manager and the third winning principle information, then the first agent determines that its role does not include the role of arbitration manager.

[0163] The above describes the collaborative method provided in this application. It should be noted that when the ACN-C network element instructs the first intelligent agent to collaborate with other intelligent agents to complete the first task via the first line of information, the collaborative method can be either centralized or distributed, which can be implemented in different ways.

[0164] For example, in one implementation, the coordination method information includes X bits. When these X bits take a first value, it indicates a centralized coordination method. When they take a second value, it indicates a distributed coordination method. For instance, the X bits can be a single bit; a value of 0 indicates a centralized coordination method, a value of 1 indicates a distributed coordination method, or vice versa.

[0165] For example, in another implementation, the role information in the first action information implicitly indicates whether a centralized or distributed collaboration method is used. That is, the role information, in addition to serving as a tool for the first agent to determine its role, also indicates the collaboration method employed. For instance, if the role information includes information instructing the first agent to assume the role of leader or includes instructions to recommend a leader, this is also considered an implicit indication of a centralized collaboration method. Similarly, if the role information includes information instructing the first agent to assume the role of a speaking manager or includes instructions to recommend a speaking manager, or if the role information includes information instructing the first agent to assume the role of an arbitration manager or includes instructions to recommend an arbitration manager, this is also considered an implicit indication of a distributed collaboration method.

[0166] Below, for ease of understanding, we will combine... Figures 4-6 This section provides two examples illustrating how ACN-C network elements are determined using a centralized collaborative approach.

[0167] Figure 4 This is a flowchart illustrating a centralized collaborative approach to complete a first task, as provided in one embodiment of this application. Figure 4 As shown, the method includes:

[0168] Step 1: Agent 1 sends capability information 1 to the ACN-C network element. Capability information 1 is used to indicate the capabilities of Agent 1.

[0169] For example, capability information 1 includes the identity (ID) of agent 1. In one implementation, capability information 1 is carried in registration message 1 sent by agent 1 to the ACN-C network element. Registration message 1 is used by agent 1 to request registration with the ACN-C network element.

[0170] The capabilities of Agent 1 include Agent 1's ontological capabilities and / or model capabilities. For example, Agent 1's ontological capabilities include one or more of the following: Agent 1's model type, operational capabilities, movement capabilities, perception capabilities, etc. For example, Agent 1's model capabilities include one or more of the following: whether it supports LLM, the type of LLM, the longest token length supported by LLM, and / or the data modality type. For a more detailed description of ontological and model capabilities, please refer to [reference needed]. Figure 3The descriptions in the embodiments are not repeated here.

[0171] Optionally, capability information 1 can also indicate whether agent 1 is capable of assuming the leader role. Indicating whether agent 1 is capable of assuming the leader role can be implemented in two ways:

[0172] 1) Displays an indication of whether one is capable of assuming the role of leader. For example, capability information 1 includes an X1 bit, which is used to indicate whether one is capable of assuming the role of leader. For example, a value of #1 indicates that one is capable of assuming the role of leader.

[0173] 2) Implicit indications of leadership ability: For example, capability information 1 includes parameters that determine whether agent 1 is capable of assuming a leadership role. These parameters might include one or more of the following: the type of LLM, the number of parameters in the LLM, and the LLM's personality traits. For instance, the main personality traits of an LLM are language comprehension and generation capabilities. Here are some characteristics of LLM models: 1) Strong language comprehension: LLMs can understand complex language structures and semantics, including vocabulary, grammar, syntax, and context. They can learn from large amounts of text data and master language usage rules and patterns. 2) Strong language generation: LLMs can generate fluent, coherent, and meaningful text based on input prompts. They can engage in multi-turn dialogues, answer questions, write articles, and even create poems and stories. 3) High adaptability: LLMs can adjust their output style and content according to different application scenarios and needs. For example, they can switch freely between formal and informal settings or generate different texts based on user preferences. 4) High plasticity: The personality traits of LLMs are not fixed but can be changed through training and fine-tuning. By adjusting the model parameters and training data, the model can exhibit different personality traits. 5) High consistency: Once LLMs are assigned certain personality traits, they will maintain consistent performance when generating text. For example, if a model is set to have an optimistic and cheerful personality, it will frequently produce positive, humorous, and witty content in the generated text.

[0174] Step 2: Agent 2 sends capability information 2 to the ACN-C network element. Capability information 2 is used to indicate the capabilities of agent 2.

[0175] For example, capability information 2 includes the identity (ID) of agent 2. In one implementation, capability information 2 is carried in registration message 2 sent by agent 2 to the ACN-C network element. Registration message 2 is used by agent 2 to request registration with the ACN-C network element.

[0176] The capabilities of Agent 2 include Agent 2's ontological capabilities and / or model capabilities. For example, Agent 2's ontological capabilities include one or more of the following: Agent 2's model, type, operational capabilities, movement capabilities, perception capabilities, etc. For example, Agent 2's model capabilities include one or more of the following: whether it supports LLM, the type of LLM, the longest token length supported by LLM, and / or the data modality type.

[0177] Optionally, capability information 2 can also indicate whether agent 2 is capable of assuming the role of leader. A detailed description of this part can be found in S401, and will not be repeated here.

[0178] Step 3: Agent 3 sends capability information 3 to the ACN-C network element. Capability information 3 is used to indicate the capabilities of agent 3.

[0179] For example, capability information 3 includes the identity (ID) of agent 3. In one implementation, capability information 3 is carried in registration message 3 sent by agent 3 to the ACN-C network element. Registration message 3 is used by agent 3 to request registration with the ACN-C network element.

[0180] The capabilities of Agent 3 include its ontological capabilities and / or model capabilities. For example, Agent 3's ontological capabilities include one or more of the following: Agent 3's model, type, operational capabilities, movement capabilities, perception capabilities, etc. For example, Agent 3's model capabilities include one or more of the following: whether it supports LLM, the type of LLM, the longest token length supported by LLM, and / or the data modality type.

[0181] Optionally, capability information 3 can also indicate whether agent 3 is capable of assuming the role of leader. A detailed description of this part can be found in S401, and will not be repeated here.

[0182] It should be noted that this application does not restrict the order in which steps 1 to 3 are performed.

[0183] Step 4: Agent 1 sends request information 1 to the ACN-C network element. Request information 1 includes information for instructing the first task.

[0184] Understandably, request information 1 can be interpreted as the first information in S310. The information used to indicate the first task is the task description information.

[0185] Optionally, request information 1 includes the identifiers of candidate agents desired to collaborate on the first task. That is, request information 1 can also be understood as indicating the candidate agents desired by agent 1. Alternatively, request information 1 can also be understood as indicating the candidate team members requested by agent 1.

[0186] Optionally, request information 1 may also include collaboration method request information, which indicates that agent 1 expects to use a centralized collaboration method to collaboratively complete the first task. Therefore, in this example, collaboration method request information can also be called centralized collaboration method request information.

[0187] Optionally, request information 1 may also include leader instruction information, which is used to instruct agent 1 to assume the role of leader.

[0188] Step 5: The ACN-C network element determines the intelligent agents (team members) to participate in completing the first task based on the first task.

[0189] In one implementation, the ACN-C network element determines the team members needed to complete the first task based on the capabilities of each agent registered with the ACN-C network element. In this example, the agents determined by the ACN-C network element to participate in completing the first task include agent 1, agent 2, and agent 3.

[0190] Optionally, if request information 1 also indicates the desired candidate agent, the ACN-C network element determines the agent to participate in completing the first task based on the candidate agent desired by agent 1.

[0191] Step 6: The ACN-C network element determines the collaboration mode, adopting a centralized collaboration mode and the roles of each member (leader or member).

[0192] Specifically, the ACN-C network element determines to adopt a centralized collaboration mode when the following conditions are met: there is an agent among the team members whose capabilities meet the requirements for assuming the Leader role in the first task, and the agent's contract also allows it to assume the Leader role.

[0193] In this example, the ACN-C network element adopts a centralized planning approach based on the primary task to be performed and the capabilities of the identified team members.

[0194] Optionally, in step 4, if agent 1's request message 1 further indicates that agent 1 expects to assume the leader role and agent 1 meets the above requirements, then ACN-C determines that agent 1 assumes the leader role. Otherwise, ACN-C determines that any agent among the team members who meets the above requirements assumes the leader role.

[0195] Step 7: The ACN-C network element interacts with the ACN-U network element, and the ACN-U network element is configured with message routing for team members.

[0196] Optionally, the ACN-C network element creates agent 4 within the ACN-U network element and determines that agent 4 is among the team members participating in completing the first task. Agent 4 is used for message monitoring among team members, and its role is that of a monitor. In essence, agent 4 is also known as a network agent.

[0197] Step 8: The ACN-C network element sends behavioral information 1 to agent 1.

[0198] In this application, behavioral information 1 can also be referred to as team response information.

[0199] Behavioral information 1 can indicate one or more of the following: team member identifiers, capability information, message template information, collaboration method information, and agent 1's role information. Capability information indicates the capabilities of agents 2, 3, and 4 (optional). Message template information indicates the message template used by agent 1 when communicating with other team members. Collaboration method information indicates that agent 1 and other team members should use a centralized collaboration method to complete the first task; therefore, in this example, the collaboration method information is also called centralized collaboration method indication information.

[0200] In this example, the role information of agent 1 includes information indicating the role that agent 1 undertakes.

[0201] Understandably, after receiving behavioral information 1, agent 1, based on the indicated message template, collaborates with agents 2 and 3 in a centralized manner to complete the first task.

[0202] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 1 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0203] Step 9: The ACN-C network element sends behavioral information 2 to agent 2.

[0204] Behavioral information 2 may indicate one or more of the following: identification information, capability information, message template information, collaboration method information, and agent 2's role information. The identification information includes the identifiers of agent 1, agent 2, and agent 3. The capability information indicates the capabilities of agent 1, agent 2, and agent 3. The message template information indicates the message template used by agent 2 when communicating with agent 1 and agent 3. The collaboration method information indicates that agent 2, agent 1, and agent 3 will use a centralized collaboration method to complete the first task; agent 1's role information includes information indicating the role undertaken by agent 2.

[0205] Understandably, after receiving behavioral information 2, agent 2, based on the indicated message template, collaborates with agent 1 and agent 3 in a centralized manner to complete the first task.

[0206] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 2 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0207] Step 10: The ACN-C network element sends behavioral information 3 to the intelligent agent 3.

[0208] Behavioral information 3 may indicate one or more of the following: identification information, capability information, message template information, collaboration method information, and role information of agent 3. The identification information includes the identifiers of agent 1, agent 2, and agent 3. The capability information indicates the capabilities of agent 1, agent 2, and agent 3. The message template information indicates the message template used by agent 3 when communicating with agent 1 and agent 2. The collaboration method information indicates that agent 3, agent 1, and agent 2 will use a centralized collaboration method to complete the first task; the role information of agent 3 includes information indicating the role undertaken by agent 3.

[0209] Understandably, after receiving behavioral information 3, agent 3, based on the indicated message template, collaborates with agent 1 and agent 2 in a centralized manner to complete the first task.

[0210] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 3 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0211] It should be noted that this application does not restrict the order in which steps 1 to 3 are performed.

[0212] Figure 5 This is a flowchart illustrating a centralized collaborative approach to complete a first task, as provided in one embodiment of this application. Figure 5 As shown, the method includes:

[0213] Step 1: Agent 1 sends capability information 1 to the ACN-C network element. Capability information 1 is used to indicate the capabilities of Agent 1.

[0214] Step 2: Agent 2 sends capability information 2 to the ACN-C network element. Capability information 2 is used to indicate the capabilities of agent 2.

[0215] Step 3: Agent 3 sends capability information 3 to the ACN-C network element. Capability information 3 is used to indicate the capabilities of agent 3.

[0216] Step 4: Agent 1 sends request information 1 to the ACN-C network element. Request information 1 includes information for instructing the first task.

[0217] Step 5: The ACN-C network element determines the intelligent agents (team members) to participate in completing the first task based on the first task.

[0218] For a detailed description of S501 to S505, please refer to Figure 4 The descriptions in S401 to S405 of the embodiments will not be repeated here.

[0219] Step 6: The ACN-C network element determines the collaboration mode and adopts a centralized collaboration mode.

[0220] How to determine the implementation method of centralized collaboration (refer to) Figure 4 The description in S406 of the embodiment will not be repeated here.

[0221] Step 7: The ACN-C network element interacts with the ACN-U and is configured in the ACN-U for message routing among team members.

[0222] This step can be implemented by referring to... Figure 4 The description of S407 in the embodiment will not be repeated here.

[0223] Step 8: The ACN-C network element sends behavioral information 1 to agent 1.

[0224] Behavioral information 1 can indicate one or more of the following: identification information, capability information, message template information, collaboration method information, and agent 1's role information. Among these, identification information, capability information, message template information, and collaboration method information are related to... Figure 4 The implementation methods are the same and will not be repeated.

[0225] This example is similar to Figure 4The difference in this embodiment lies in the role information of agent 1. Specifically, in this example, the role information of agent 1 includes leader recommendation instruction information. Optionally, behavior information 1 also includes first winning principle information. The leader recommendation instruction information is used to instruct the execution of the leader recommendation process, and the first winning principle information is used to determine the agent among the team members who assumes the leader role. For example, the first winning principle information includes: the one with the most votes wins; if the number of votes is the same, the one who spoke first (the agent recommended first) wins.

[0226] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 1 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0227] Step 9: The ACN-C network element sends behavioral information 2 to agent 2.

[0228] Behavioral information 2 can indicate one or more of the following: identification information, capability information, message template information, collaboration method information, and agent 1's role information. Among these, identification information, capability information, message template information, and collaboration method information are related to... Figure 4 The implementation methods are the same and will not be repeated.

[0229] This example is similar to Figure 4 The difference in this embodiment lies in the role information of agent 2. Specifically, in this example, the role information of agent 2 includes recommended leader instruction information. Optionally, the behavior information 2 also includes first winning principle information. The recommended leader instruction information and the first winning principle information can be referred to the description in step 8, and will not be repeated here.

[0230] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 2 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0231] Step 10: The ACN-C network element sends behavioral information 3 to the intelligent agent 3.

[0232] Behavioral information 3 can indicate one or more of the following: identification information, capability information, message template information, collaboration method information, and agent 1's role information. Among these, identification information, capability information, message template information, and collaboration method information are related to... Figure 4 The implementation methods are the same and will not be repeated.

[0233] This example is similar to Figure 4 The difference in this embodiment lies in the role information of agent 3. Specifically, in this example, the role information of agent 3 includes recommended leader instruction information. Behavioral information 3 also includes first winning principle information. The recommended leader instruction information and the first winning principle information can be found in the description in step 8, and will not be repeated here.

[0234] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 3 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0235] Step 11: Agent 1 determines the recommended agent to assume the leader role.

[0236] Step 12: Agent 1 sends Recommendation Message 1 to the ACN-U network element. Recommendation Message 1 includes the agent recommended by Agent 1 to assume the leader role.

[0237] Step 13: The ACN-U network element sends recommendation message 1 to agent 2.

[0238] Step 14: The ACN-U network element sends recommendation message 1 to agent 3.

[0239] Step 15: Agent 2 determines the recommended agent to assume the leader role.

[0240] Step 16: Agent 2 sends recommendation message 2 to the ACN-U network element. Recommendation message 2 includes the agent recommended by Agent 2 to assume the leader role.

[0241] Step 17: The ACN-U network element sends recommendation message 2 to agent 3.

[0242] Step 18: The ACN-U network element sends recommendation message 2 to agent 1.

[0243] Step 19: Agent 3 determines the recommended agent to assume the leader role.

[0244] Step 20: Agent 3 sends recommendation message 3 to the ACN-U network element. Recommendation message 3 includes the agent recommended by Agent 3 to assume the leader role.

[0245] Step 21: The ACN-U network element sends recommendation message 3 to agent 1.

[0246] Step 22: The ACN-U network element sends recommendation message 3 to agent 2.

[0247] Step 23: Agent 1 determines its role based on the first winning principle information.

[0248] Step 24: Agent 2 determines its role based on the first winning principle information.

[0249] Step 25: Agent 3 determines its role based on the first winning principle information.

[0250] It can be seen that, Figure 5 The example shown is the same as Figure 4 The difference in the example shown is that the ACN-C network element does not determine the roles of agent 1, agent 2 and agent 3 on its own. Instead, the roles of agent 1, agent 2 and agent 3 are determined by themselves based on the process of recommending a leader.

[0251] refer to Figure 6 , Figure 6 This is a flowchart illustrating the process of an intelligent agent 1, acting as a leader, instructing one of its members to perform a sub-task, according to one embodiment of this application. Figure 6 As shown, the method includes:

[0252] Step 1: Agent 1 determines the subtasks assigned to Agent 2 based on the first task.

[0253] Step 2: Agent 1 sends task allocation information to Agent 2 through the ACN-U network element.

[0254] For example, the task allocation information includes the identifier of agent 2 and semantic instructions, with the semantic instructions indicating the sub-tasks to be executed by agent 2. Correspondingly, after receiving the task allocation information, the ACN-U network element sends the task allocation information to agent 2.

[0255] Step 3: The ACN-U network element forwards the corresponding task allocation information to agent 2.

[0256] Optionally, before step 3, the ACN-U network element can perform step 2.a: verify the task allocation information corresponding to agent 2. Further, if the verification is successful, the ACN-U network element forwards the semantic instructions corresponding to agent 2 to agent 2.

[0257] Step 4: Agent 2 executes the subtask indicated by the semantic instruction.

[0258] Step 5: After completing the subtask indicated by the semantic instruction, agent 2 sends a response message to the ACN-U network element. The response message is used to indicate the completion of the subtask assigned to agent 2.

[0259] Step 6: The ACN-U network element forwards the response message sent by agent 2 to agent 1.

[0260] Below, for ease of understanding, we will combine... Figures 7-12 This section describes two examples of how ACN-C network elements are determined using a distributed collaborative approach.

[0261] refer to Figure 7 , Figure 7 The methods shown include:

[0262] Step 1: Agent 1 sends capability information 1 to the ACN-C network element. Capability information 1 is used to indicate the capabilities of Agent 1.

[0263] For example, capability information 1 includes the identity (ID) of agent 1. In one implementation, capability information 1 is carried in registration message 1 sent by agent 1 to ACN-C, and registration message 1 is used by agent 1 to request registration with ACN-C network element.

[0264] For example, the ontological capabilities of Agent 1 include one or more of the following: Agent 1's model, type, operational capabilities, motor capabilities, and perception capabilities. For example, the model capabilities of Agent 1 include one or more of the following: whether it supports LLM, the type of LLM, the longest token length supported by LLM, and / or the data modality type. More detailed descriptions of ontological and model capabilities can be found in [reference needed]. Figure 3 The descriptions in the embodiments are not repeated here.

[0265] Optionally, capability information 1 can also indicate whether agent 1 is capable of assuming the roles of chartmanager and / or arbitrator.

[0266] For example, indicating whether agent 1 is capable of assuming the role of a chart manager can be implemented in two ways: 1) Explicitly indicating whether it is capable of assuming the role of a chart manager. For example, capability information 1 includes 1 bit, which is used to indicate whether it is capable of assuming the role of a chart manager. For example, a value of 1 indicates capability, a value of 0 indicates inability, or vice versa. 2) Implicitly indicating whether it is capable of assuming the role of a chart manager. For example, capability information 1 includes some parameters, based on which it can be determined whether agent 1 is capable of assuming the role. For example, the parameters used to determine whether agent 1 is capable of assuming the role of chart manager include one or more of the following: the type of LLM, the number of parameters of the LLM, the personality traits of the LLM, etc.

[0267] For example, indicating whether agent 1 is capable of assuming the role of arbitration manager can be implemented in two ways: 1) Explicitly indicating whether it is capable of assuming the role of arbitration manager. For example, capability information 1 includes 1 bit, which is used to indicate whether it is capable of assuming the role of arbitration manager. For example, a value of 1 indicates capability, a value of 0 indicates inability, or vice versa. 2) Implicitly indicating whether it is capable of assuming the role of arbitration manager. For example, capability information 1 includes some parameters, based on which it can be determined whether agent 1 is capable of assuming the role. For example, the parameters used to determine whether agent 1 is capable of assuming the role of arbitration manager include one or more of the following: the type of LLM, the number of parameters of the LLM, the personality traits of the LLM, etc.

[0268] For more detailed information on LLM types, the number of parameters in an LLM, and the personality traits associated with an LLM, please refer to [link / reference needed]. Figure 3 The descriptions in the embodiments are omitted.

[0269] Step 2: Agent 2 sends capability information 2 to the ACN-C network element. Capability information 2 is used to indicate the capabilities of agent 2.

[0270] For example, capability information 2 includes the identity (ID) of agent 2. In one implementation, capability information 2 is carried in registration message 2 sent by agent 2 to ACN-C, and registration message 2 is used by agent 2 to request registration with ACN-C network elements.

[0271] For example, the ontological capabilities of Agent 1 may include one or more of the following: Agent 2's model, type, operational capabilities, motor capabilities, and perception capabilities. For example, the model capabilities of Agent 2 may include one or more of the following: whether it supports LLM, the type of LLM, the longest token length supported by LLM, and / or the data modality type. More detailed descriptions of ontological and model capabilities can be found in [reference needed]. Figure 3 The descriptions in the embodiments are not repeated here.

[0272] Optionally, capability information 2 can also indicate whether agent 2 is capable of assuming the roles of chartmanager and / or arbitrator. The implementation of this part is analogous to the implementation of capability information 1 in step 1 indicating whether agent 1 is capable of assuming the roles of chartmanager and / or arbitrator, and will not be elaborated further.

[0273] Step 3: Agent 2 sends capability information 3 to the ACN-C network element. Capability information 3 is used to indicate the agent 3's ontological capability 3 and / or model capability 3.

[0274] For example, capability information 3 includes the identity (ID) of agent 3. In one implementation, capability information 3 is carried in registration message 3 sent by agent 3 to ACN-C, and registration message 3 is used by agent 3 to request registration with ACN-C network elements.

[0275] For example, the ontological capabilities of Agent 1 may include one or more of the following: Agent 3's model type, operational capabilities, motor capabilities, and perception capabilities. For example, the model capabilities of Agent 3 may include one or more of the following: whether it supports LLM, the type of LLM, the longest token length supported by LLM, and / or the data modality type. More detailed descriptions of ontological and model capabilities can be found in [reference needed]. Figure 3The descriptions in the embodiments are not repeated here.

[0276] Optionally, capability information 3 can also indicate whether agent 3 is capable of assuming the roles of chartmanager and / or arbitrator. The implementation of this part is analogous to the implementation of capability information 1 in step 1 indicating whether agent 1 is capable of assuming the roles of chartmanager and / or arbitrator, and will not be elaborated further.

[0277] Step 4: Agent 1 sends request information 1 to the ACN-C network element. Request information 1 includes information for instructing the first task.

[0278] Understandably, request information 1 can be understood as the first piece of information.

[0279] Optionally, request information 1 includes the identifiers of candidate agents desired to collaborate on the first task. That is, request information 1 can also be understood as indicating the candidate agents desired by agent 1. Alternatively, request information 1 can also be understood as indicating the candidate team members requested by agent 1.

[0280] Optionally, request information 1 may also include cooperative mode request information, which indicates that agent 1 expects to use a distributed cooperative mode to collaboratively complete the first task. Therefore, in this example, the cooperative mode request information can also be called distributed cooperative mode request information.

[0281] Step 5: The ACN-C network element determines the intelligent agents (team members) to participate in completing the first task based on the first task.

[0282] This implementation method can be referenced. Figure 4 The description of step 5 in the embodiments will not be repeated here.

[0283] Step 6: The ACN-C network element determines the collaborative mode and adopts a distributed collaborative mode.

[0284] Specifically, the ACN-C network element determines to adopt a distributed collaborative approach when the following conditions are met: each of the at least two agents that collaboratively complete the first task has the ability to perform distributed planning and the contract also allows them to adopt a distributed collaborative approach.

[0285] Specifically, when a distributed collaborative approach is adopted, the ACN-C network element will also determine dialogue flow control information and model negotiation capability information. The meaning of dialogue flow control information can be found in [reference needed]. Figure 3 The descriptions in the embodiments are not repeated here.

[0286] Specifically, in this example, if the dialogue flow control information indicates that speaking should be performed according to the speaking manager's schedule, the ACN-C network element also identifies the speaking manager among the team members. Furthermore, if the dialogue flow control information indicates that arbitration should be performed according to the arbitration manager, the arbitration manager among the team members is also identified.

[0287] Step 7: The ACN-C network element interacts with the ACN-U network element, and the ACN-U network element is configured with message routing for team members.

[0288] Optionally, the ACN-C network element creates agent 4 in the ACN-U network element. Agent 4 is used for message supervision among team members, and the role of agent 4 is set as monitor.

[0289] Optionally, the ACN-C network element determines that agent 4 will assume the roles of speech manager and / or arbitration manager and sets the role of agent 4 as speech manager and / or arbitration manager.

[0290] Step 8: The ACN-C network element sends behavioral information 1 to agent 1.

[0291] Behavioral information 1 can also be called team response information.

[0292] Behavioral information 1 can indicate one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and agent 1's role information. The descriptions of capability information, message template information, dialogue flow control information, and model negotiation capability information can be found above and will not be repeated here.

[0293] In this example, the collaboration method information is used to instruct agent 1 to complete the first task in a distributed collaborative manner with agents 2 and 3. Therefore, the collaboration method information in this example is also called the distributed collaboration method instruction information.

[0294] In this example, the role information includes indication information used to indicate the role that agent 1 assumes.

[0295] In this embodiment, the ACN-C network element determines that the role of intelligent agent 1 includes member. Further, if the ACN-C network element determines that the role of intelligent agent 1 also includes a speech manager, then the role information also includes instructions instructing intelligent agent 1 to assume the role of speech manager. If the ACN-C network element determines that the role of intelligent agent 1 also includes an arbitration manager, then the role information also includes instructions instructing intelligent agent 1 to assume the role of arbitration manager.

[0296] It should be noted that the aforementioned behavioral information 1, which indicates that the distributed collaboration method can be implemented in different ways when the first task is completed collaboratively.

[0297] For example, in one implementation, a distributed collaborative approach is indicated by displaying an X2 bit.

[0298] For example, in another implementation, if the ACN-C network element instructs agent 1 to assume the role of a speech manager and / or arbitration manager, then the behavior information 1 may not include the aforementioned X2 bits. Correspondingly, after receiving behavior information 1, based on its assumed role as a speech manager and / or arbitration manager, agent 1 determines to use a distributed collaborative approach to cooperate with other agents to complete the first task. In other words, in this implementation, the ACN-C network element implicitly instructs agent 1 to use a distributed collaborative approach based on its instruction that agent 1 assume the role of a speech manager and / or arbitration manager.

[0299] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 1 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0300] Step 9: ACN-C sends behavioral pattern information 2 to agent 2.

[0301] Behavioral model information 2 can indicate one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and agent 2's role information. The descriptions of capability information, message template information, dialogue flow control information, and model negotiation capability information can be found above and will not be repeated here.

[0302] In this example, the collaboration method information is used to instruct agent 2 to complete the first task in a distributed collaboration manner with agent 1 and agent 3. Therefore, the collaboration method information in this example is also called the distributed collaboration method instruction information.

[0303] In this example, the role information of agent 2 includes indication information for indicating the role that agent 2 undertakes.

[0304] In this embodiment, the ACN-C network element determines that the role of intelligent agent 2 includes member. Further, if the ACN-C network element determines that the role of intelligent agent 2 also includes a speech manager, then the role information also includes instructions instructing intelligent agent 2 to assume the role of speech manager. If the ACN-C network element determines that the role of intelligent agent 2 also includes an arbitration manager, then the role information also includes instructions instructing intelligent agent 2 to assume the role of arbitration manager.

[0305] It should be noted that the above-mentioned behavioral information 2 indicates that the distributed collaboration method can be implemented in different ways when the first task is completed collaboratively.

[0306] For example, in one implementation, a distributed collaborative approach is indicated by displaying an X2 bit.

[0307] For example, in another implementation, if the ACN-C network element instructs agent 2 to assume the role of a speech manager and / or arbitration manager, then the behavior information 2 may not include the aforementioned X2 bits. Correspondingly, after receiving behavior information 2, based on its assumed role as a speech manager and / or arbitration manager, agent 2 determines to use a distributed collaborative approach to cooperate with other agents to complete the first task. In other words, in this implementation, the ACN-C network element implicitly instructs agent 2 to use a distributed collaborative approach based on its instruction that agent 2 assume the role of a speech manager and / or arbitration manager.

[0308] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 2 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0309] Step 10: ACN-C sends behavioral pattern information 3 to Agent3.

[0310] Behavioral model information 3 can indicate one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and agent 3 role information. The descriptions of capability information, message template information, dialogue flow control information, and model negotiation capability information can be found above and will not be repeated here.

[0311] In this example, the collaboration method information is used to instruct agent 3 to complete the first task in a distributed collaboration manner with agents 1 and 2. Therefore, the collaboration method information in this example is also called the distributed collaboration method instruction information.

[0312] In this example, the role information of agent 3 includes indication information for indicating the role that agent 3 undertakes.

[0313] In this embodiment, the ACN-C network element determines that the role of intelligent agent 3 includes member. Further, if the ACN-C network element determines that the role of intelligent agent 3 also includes a speech manager, then the role information also includes instructions instructing intelligent agent 3 to assume the role of speech manager. If the ACN-C network element determines that the role of intelligent agent 3 also includes an arbitration manager, then the role information also includes instructions instructing intelligent agent 3 to assume the role of arbitration manager.

[0314] It should be noted that the above-mentioned behavioral information 3 indicates that the distributed collaboration method can be implemented in different ways when the first task is completed collaboratively.

[0315] For example, in one implementation, a distributed collaborative approach is indicated by displaying an X2 bit.

[0316] For example, in another implementation, if the ACN-C network element instructs agent 3 to assume the role of a speech manager and / or arbitration manager, then the behavior information 3 may not include the aforementioned X2 bits. Correspondingly, after receiving behavior information 3, agent 3, based on its assumed role as a speech manager and / or arbitration manager, determines to use a distributed collaborative approach to cooperate with other agents to complete the first task. In other words, in this implementation, the ACN-C network element implicitly instructs agent 3 to use a distributed collaborative approach based on its instruction that agent 3 assume the role of a speech manager and / or arbitration manager.

[0317] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 3 may also indicate one or more of the following: the identifier of agent 4, the capabilities of agent 4.

[0318] Furthermore, agent 1 assigns the first task to agents 2 and 3.

[0319] It can be seen that, Figure 7 In the example shown, after the ACN-C network element determines the distributed collaboration method and team members, it assigns the roles of each team member to each team member, so that the team members can cooperate with each other based on the instructions to complete the first task.

[0320] refer to Figure 8 , Figure 8 The methods shown include:

[0321] Step 1: Agent 1 sends capability information 1 to the ACN-C network element. Capability information 1 is used to indicate the capabilities of Agent 1.

[0322] Step 2: Agent 2 sends capability information 2 to the ACN-C network element. Capability information 2 is used to indicate the capabilities of agent 2.

[0323] Step 3: Agent 2 sends capability information 3 to ACN-C. Capability information 3 is used to indicate the agent 3's ontological capability 3 and / or model capability 3.

[0324] Step 4: Agent 1 sends request information 1 to the ACN-C network element. Request information 1 includes information for instructing the first task.

[0325] Step 5: The ACN-C network element determines the intelligent agents (team members) to participate in completing the first task based on the first task.

[0326] Step 6: The ACN-C network element determines the collaborative mode and adopts a distributed collaborative mode.

[0327] For details on how ACN-C network elements determine the distributed collaborative approach, please refer to [link to relevant documentation]. Figure 7 The descriptions in the embodiments are omitted.

[0328] Specifically, when a distributed collaborative approach is adopted, the ACN-C network element will also determine dialogue flow control information and model negotiation capability information. The meaning of dialogue flow control information can be found in [reference needed]. Figure 3 The descriptions in the embodiments are not repeated here.

[0329] Specifically, in this example, the dialogue flow control information indicates that speaking should be performed according to the speaking manager's schedule, and / or, the dialogue flow control information indicates that arbitration should be performed according to the arbitration manager.

[0330] Step 7: The ACN-C network element interacts with the ACN-U network element, and the ACN-U network element is configured with message routing for team members.

[0331] For a detailed description of this section, please refer to [link / reference]. Figure 7 The descriptions in the embodiments are omitted.

[0332] Step 8: The ACN-C network element sends behavioral information 1 to agent 1.

[0333] Behavioral information 1 can also be called team response information.

[0334] Behavioral information 1 may indicate one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and agent 1's role information. Among these, the descriptions of capability information, message template information, dialogue flow control information, model negotiation capability information, and collaboration method information can be... Figure 7 The description of step 8 in the embodiment will not be repeated here.

[0335] In this example, if the dialogue flow control information indicates that speaking should be done according to the speaking manager's schedule, the role information includes a recommended speaking manager instruction. This instruction instructs agent 1 to determine its role based on the recommended speaking manager's process. Optionally, behavior information 1 also includes a second winning rule, allowing agent 1 to determine the speaking manager based on the second winning rule, thereby further determining its own role. And / or, if the dialogue flow control information indicates that arbitration should be conducted based on an arbitration manager, the role information includes a recommended arbitration manager instruction. This instruction instructs agent 1 to determine its role based on the recommended arbitration manager's process. Optionally, behavior information 1 also includes a third winning rule, allowing agent 1 to determine the arbitration manager based on the third winning rule, thereby further determining its own role.

[0336] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, then behavioral information 1 may also indicate one or more of the following: the identifier of agent 4, and the capabilities of agent 4. See details for further information. Figure 7 The descriptions in the embodiments are omitted.

[0337] Step 9: The ACN-C network element sends behavioral information 2 to agent 2.

[0338] Behavioral information 2 can indicate one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and agent 2's role information. The descriptions of capability information, message template information, dialogue flow control information, model negotiation capability information, and collaboration method information can be found in [reference needed]. Figure 7 The description of step 8 in the embodiment will not be repeated here.

[0339] In this example, if the dialogue flow control information indicates that speaking should be done according to the speaking manager's schedule, the role information of agent 2 includes a recommended speaking manager instruction. This instruction instructs agent 2 to determine its role based on the recommended speaking manager process. Optionally, the behavior information 2 also includes a second winning rule, allowing agent 2 to determine the speaking manager based on the second winning rule, thereby further determining its own role. And / or, if the dialogue flow control information indicates that arbitration should be conducted based on an arbitration manager, the role information of agent 2 includes a recommended arbitration manager instruction. This instruction instructs agent 2 to determine its role based on the recommended arbitration manager process. Optionally, the behavior information 2 also includes a third winning rule, allowing agent 2 to determine the arbitration manager based on the third winning rule, thereby further determining its own role.

[0340] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 2 may also indicate one or more of the following: the identifier of agent 4, and the capabilities of agent 4. See details for further information. Figure 7 The descriptions in the embodiments are omitted.

[0341] Step 10: The ACN-C network element sends behavioral information 3 to the intelligent agent 3.

[0342] Behavioral information 3 can indicate one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and agent 3 role information. The descriptions of capability information, message template information, dialogue flow control information, model negotiation capability information, and collaboration method information can be found in [reference needed]. Figure 7The description of step 8 in the embodiment will not be repeated here.

[0343] In this example, if the dialogue flow control information indicates that speaking should be done according to the speaking manager's schedule, the role information of agent 3 includes a recommended speaking manager instruction. This instruction instructs agent 3 to determine its role based on the recommended speaking manager's process. Optionally, the behavior information 3 also includes a second winning rule, allowing agent 3 to determine the speaking manager based on the second winning rule, thereby further determining its own role. And / or, if the dialogue flow control information indicates that arbitration should be conducted based on an arbitration manager, the role information of agent 3 includes a recommended arbitration manager instruction. This instruction instructs agent 3 to determine its role based on the recommended arbitration manager's process. Optionally, the behavior information 3 also includes a third winning rule, allowing agent 3 to determine the arbitration manager based on the third winning rule information, thereby further determining its own role.

[0344] Optionally, if the ACN-C network element creates agent 4 in the ACN-U network element, the behavioral information 3 may also indicate one or more of the following: the identifier of agent 4, and the capabilities of agent 4. See details for further information. Figure 7 The descriptions in the embodiments are omitted.

[0345] Step 11, Agent 1 determines the recommended speech manager and / or arbitration manager.

[0346] Step 12: Agent 1 sends Recommendation Message 1 to the ACN-U network element. Recommendation Message 1 includes the speech manager and / or arbitration manager recommended by Agent 1.

[0347] Step 13: Send recommendation message 1 to agent 2.

[0348] Step 14: Send recommendation message 1 to agent 3.

[0349] Step 15, Agent 2 determines the speaking manager and / or arbitration manager.

[0350] Step 16: Agent 2 sends recommendation message 2, which includes the speech manager and / or arbitration manager recommended by Agent 2.

[0351] Step 17: Send recommendation message 2 to agent 3.

[0352] Step 18: Send recommendation message 2 to agent 1.

[0353] Step 19, Agent 3 determines the recommended speech manager and / or arbitration manager.

[0354] Step 20: Agent 3 sends recommendation message 3, which includes the speech manager and / or arbitration manager recommended by Agent 3.

[0355] Step 21: Send recommendation message 3 to agent 1.

[0356] Step 22: Send recommendation message 3 to agent 2.

[0357] Step 23: Agent 1 determines its role according to the second winning rule and / or the third winning rule.

[0358] Step 24: Agent 2 determines its role according to the second and / or third winning rules.

[0359] Step 25: Agent 3 determines its role according to the second and / or third winning rules.

[0360] It can be seen that, Figure 8 In the example shown, after the ACN-C network element determines the distributed collaboration method and team members, it sends recommended speaking managers and / or recommended arbitration managers to each team member, enabling each team member to determine their own role and collaborate with other team members based on their respective roles to complete the first task.

[0361] refer to Figure 9 , Figure 9 The diagram illustrates the process of team members speaking in a fixed order, using a network agent (e.g., agent 4) created within the ACN-U network element as a summary of their remarks. Figure 9 As shown, the method includes:

[0362] Step 1: Agent 1 makes decision 1 based on the first task.

[0363] Step 2: Agent 1 speaks according to the fixed speaking order indicated by the ACN-C network element and sends negotiation message 1 to the ACN-U network element. Negotiation message 1 includes decision 1 and environmental perception information 1.

[0364] Step 3: The ACN-U network element sends negotiation message 1 to agent 2.

[0365] Step 4: The ACN-U network element sends negotiation message 1 to agent 3.

[0366] Step 5: Agent 2 formulates decision 2 based on the first task.

[0367] Step 6: Agent 2 speaks according to the fixed speaking order indicated by the ACN-C network element and sends negotiation message 2 to the ACN-U network element. Negotiation message 2 includes decision 2 and environmental perception information 2.

[0368] Step 7: The ACN-U network element sends negotiation message 2 to agent 1.

[0369] Step 8: The ACN-U network element sends negotiation message 2 to agent 3.

[0370] Step 9: Agent 3 makes decision 3 based on the first task.

[0371] Step 10: Agent 3 speaks according to the fixed speaking order indicated by the ACN-C network element and sends negotiation message 3 to the ACN-U network element. The negotiation message 3 includes decision 3 and environmental perception information 3.

[0372] Step 11: The ACN-U network element sends negotiation message 3 to agent 1.

[0373] Step 12: The ACN-U network element sends negotiation message 3 to agent 2.

[0374] Step 13: Agent 4 summarizes the statements of each agent to form a negotiation summary.

[0375] Step 14: Agent 4 sends a negotiation summary to Agent 1. Agent 1 returns a confirmation message after successfully verifying the negotiation summary.

[0376] Step 15: Agent 4 sends a negotiation summary to Agent 2. Agent 2 returns a confirmation message after successfully verifying the negotiation summary.

[0377] Step 16: Agent 4 sends a negotiation summary to Agent 3. Agent 3 returns a confirmation message after successfully verifying the negotiation summary.

[0378] Step 17: Agent 4 confirms execution and instructs Agent 1, Agent 2 and Agent 3 to execute according to their respective confirmed decisions.

[0379] refer to Figure 10 , Figure 10 The summary speech is given by a network agent (e.g., agent 4) created in the ACN-U network element, illustrating the process of the team speaking in a dynamic order. For example... Figure 10 As shown, the method includes:

[0380] Step 1: Agent 4 determines that the next speaker is Agent 1.

[0381] Step 2: Agent 4 instructs Agent 1 to speak.

[0382] Step 3: Agent 1 makes decision 1 based on the first task.

[0383] Step 4: Agent 1 sends negotiation message 1 to the ACN-U network element. Negotiation message 1 includes decision 1 and environmental perception information 1.

[0384] Step 5: The ACN-U network element sends a negotiation request message 1 to agent 2.

[0385] Step 6: The ACN-U network element sends a negotiation request message 1 to agent 2.

[0386] Step 7: Agent 4 determines that the next speaker is Agent 2.

[0387] Step 8: Agent 4 instructs Agent 2 to speak.

[0388] Step 9: Agent 2 makes decision 2 based on the first task.

[0389] Step 10: Agent 2 sends negotiation message 2 to the ACN-U network element. Negotiation message 2 includes decision 2 and environmental perception information 2.

[0390] Step 11: The ACN-U network element sends a negotiation request message 2 to agent 1.

[0391] Step 12: The ACN-U network element sends a negotiation request message 2 to agent 3.

[0392] Step 13: Agent 4 determines that the next speaker is Agent 3.

[0393] Step 14: Agent 4 instructs Agent 3 to speak.

[0394] Step 15: Agent 3 makes decision 3 based on the first task.

[0395] Step 16: Agent 3 sends negotiation message 3 to the ACN-U network element. Negotiation message 3 includes decision 3 and environmental perception information 3.

[0396] Step 17: The ACN-U network element sends negotiation message 3 to agent 1.

[0397] Step 18: The ACN-U network element sends negotiation message 3 to agent 2.

[0398] Step 19: Agent 4 summarizes the statements of each agent to form a negotiation summary;

[0399] Step 20: Agent 4 sends a negotiation summary to Agent 1. Agent 1 returns a confirmation message after successfully verifying the negotiation summary.

[0400] Step 21: Agent 4 sends a negotiation summary to Agent 2. Agent 2 returns a confirmation message after successfully verifying the negotiation summary.

[0401] Step 22: Agent 4 sends a negotiation summary to Agent 3. Agent 3 returns a confirmation message after successfully verifying the negotiation summary.

[0402] Step 23: Agent 4 confirms execution and instructs Agent 1, Agent 2 and Agent 3 to execute according to their respective confirmed decisions.

[0403] refer to Figure 11 , Figure 11 Taking a team consisting of Agent 1, Agent 2, Agent 3, and Agent 4 (configured within the ACN-U network element of the ACN-C network element) as an example, this illustrates the flowchart of arbitration based on agent priority. Specifically, in this example, Agent 1 has a higher priority than Agent 3.

[0404] like Figure 11 As shown, the method includes:

[0405] Step 1: Agent 3 determines that its decision 3 conflicts with Agent 1's decision 1 and re-formulates decision 3'.

[0406] For example, in one implementation, agent 3 determines that decision 3 conflicts with decision 1 of agent 1.

[0407] For example, in another implementation, after receiving the indication information from the ACN-U network element that there is a conflict between decision 1 of agent 1 and decision 3 of agent 3 (step 1.2 in the figure), agent 3 determines that its own decision 3 conflicts with decision 1 of agent 1.

[0408] Optionally, after receiving the indication information from agent 1 that agent 1's decision 1 and agent 3's decision 3 conflict (step 1.1 in the figure), the ACN-U network element indicates to agent 3 that agent 1's decision 1 and agent 3's decision 3 conflict.

[0409] Optionally, the ACN-U network element indicates to agent 2 that agent 1's decision 1 and agent 3's decision 3 are in conflict (step 1.3 in the figure).

[0410] Step 2: Agent 3 sends negotiation message 4 to ACN-U network element. Negotiation message 4 includes decision 3' and environmental perception information 3.

[0411] Step 3: The ACN-U network element sends a negotiation message 4 to agent 1.

[0412] Optionally, the ACN-U network element can also send negotiation message 4 to agent 2.

[0413] Step 4: Agent 4 summarizes the statements of each agent and re-forms a negotiation summary;

[0414] Step 5: Agent 4 sends a negotiation summary to Agent 1. Agent 1 returns a confirmation message after successfully verifying the negotiation summary.

[0415] Step 6: Agent 4 sends a negotiation summary to Agent 2. Agent 2 returns a confirmation message after successfully verifying the negotiation summary.

[0416] Step 7: Agent 4 sends a negotiation summary to Agent 3. Agent 3 returns a confirmation message after successfully verifying the negotiation summary.

[0417] Step 8: Agent 4 confirms execution and instructs Agent 1, Agent 2 and Agent 3 to execute according to their respective confirmed decisions.

[0418] Optionally, if a conflict still exists between decision 1 and decision 3', agent 3 continues to reformulate the decision until there is no conflict between the decisions of agent 1, agent 2 and agent 3 or the negotiation depth is reached and then the negotiation stops.

[0419] refer to Figure 12 , Figure 12 Taking a team consisting of Agent 1, Agent 2, Agent 3, and Agent 4 (configured within an ACN-U network element from an ACN-C network element) as an example, this illustrates a flowchart of conflict management by an arbitration manager. In this example, Agent 4 serves as the arbitration manager. Figure 12 As shown, the method includes:

[0420] Step 1: Agent 4 determines that Agent 1's decision 1 and Agent 3's decision 3 are in conflict. Agent 4 arbitrates and determines that Agent 3 should make a new decision.

[0421] This application does not restrict how agent 4 determines the conflict between decision 1 of agent 1 and decision 3 of agent 3.

[0422] For example, in one implementation, such as Figure 12 As shown in step 1.1, agent 1 indicates to agent 4 that agent 1's decision 1 and agent 3's decision 3 conflict. Correspondingly, agent 4 learns of the conflict between agent 1's decision 1 and agent 3's decision 3 based on agent 1's instruction.

[0423] For example, in another implementation, such as Figure 12 As shown in step 1.2, agent 3 indicates to agent 1 that agent 1's decision 1 and agent 3's decision 3 conflict. Correspondingly, agent 4 learns of the conflict between agent 1's decision 1 and agent 3's decision 3 based on agent 3's instruction.

[0424] For example, in another implementation, agent 4 determines that there is a conflict between decision 1 of agent 1 and decision 3 of agent 3.

[0425] Step 2: Agent 4 sends a re-decision instruction to Agent 3, which instructs Agent 3 to make a new decision.

[0426] Step 3, agent 3 re-formulates decision 3'.

[0427] Step 4: Agent 3 sends negotiation message 4 to the ACN-U network element. The negotiation message 4 includes decision 3' and environmental perception information 3.

[0428] Step 5: The ACN-U network element sends a negotiation request message 4 to agent 1.

[0429] Optionally, the ACN-U network element sends a negotiation request message 4 to agent 2.

[0430] Step 6: Agent 4 summarizes the statements of the agents and re-forms a negotiation summary.

[0431] Step 7: Agent 4 sends a negotiation summary to Agent 1. Agent 1 returns a confirmation message after successfully verifying the negotiation summary.

[0432] Step 8: Agent 4 sends a negotiation summary to Agent 2. Agent 2 returns a confirmation message after successfully verifying the negotiation summary.

[0433] Step 9: Agent 4 sends a negotiation summary to Agent 3. Agent 3 returns a confirmation message after successfully verifying the negotiation summary.

[0434] Step 10: Agent 4 confirms execution and instructs Agent 1, Agent 2 and Agent 3 to execute according to their respective confirmed decisions.

[0435] Optionally, if agent 4 determines that there is still a conflict between decision 1 and decision 3, it continues to instruct agent 3 to reformulate the decision until there is no conflict between the decisions of agent 1, agent 2 and agent 3 or the negotiation depth is reached and then the negotiation stops.

[0436] Understandably, there may be situations where an agent's speech is abnormal. Such abnormalities include the agent not speaking in the fixed order, speaking in the wrong order, not speaking in the required order, not speaking when scheduled by the speech manager, or speaking even though not scheduled to speak.

[0437] In this application, when a speech anomaly occurs, one processing method includes: the ACN-U network element instructs the agent with the speech anomaly to no longer cooperate with other agents to complete the first task, and selects a new agent and instructs the new agent to cooperate with other agents with normal speech to complete the first task.

[0438] Below, for ease of understanding, we will combine... Figure 13Please provide an explanation.

[0439] refer to Figure 13 , Figure 13 This is a flowchart illustrating a communication method provided in one embodiment of this application. Figure 13 As shown, the method includes:

[0440] Step 1: The ACN-U network element determines that agent 2's speech is abnormal.

[0441] For example, if the agents working together to complete the first task speak in a fixed order, then agent 2 is considered to have spoken abnormally if any of the following occurs: agent 2 did not speak in the fixed order, or agent 2 spoke in the wrong order, or agent 2 did not speak in the required order.

[0442] For example, if the agents that work together to complete the first task speak based on the scheduling of ACN-U network elements, then agent 2 is determined to be speaking abnormally if any of the following situations occur: agent 2 fails to speak after the timeout period following the sending of the speaking instruction, or agent 2 speaks even though it was not scheduled to speak.

[0443] Step 2: The ACN-U network element sends a team exit message to agent 2, indicating that agent 2 will no longer cooperate with other agents to complete the first task.

[0444] Step 3: The ACN-U network element sends the corresponding behavioral information to agent 5. The behavioral information of agent 5 instructs agent 5 to cooperate with agent 1 and agent 2 to complete the first task.

[0445] The behavioral information corresponding to Agent 5 indicates one or more of the following: team member identification, capability information, message template information, collaboration method information, dialogue flow control information, model negotiation capability information, and Agent 5's role information. The descriptions of capability information, message template information, dialogue flow control information, and model negotiation capability information can be found above and will not be repeated here.

[0446] Step 4: The ACN-U network element updates the behavioral information corresponding to Agent 1 sent to Agent 1.

[0447] The updated behavioral information includes the identifier of team members, which is the identifier of agent 5.

[0448] Optionally, if the capability information, message template information, collaboration method information, role information, etc. in the behavior information corresponding to 1 are updated, then the updated capability information, message template information, collaboration method information, and role information are sent.

[0449] Step 5: The ACN-U network element updates the behavioral information corresponding to agent 3 sent to agent 3.

[0450] The updated behavioral information includes the identifier of team members, which is the identifier of agent 5.

[0451] Optionally, if the capability information, message template information, collaboration method information, role information, etc. in the behavior information corresponding to 1 are updated, then the updated capability information, message template information, collaboration method information, and role information are sent.

[0452] Understandably, in adopting Figure 13 Following the method shown, Agent 1, Agent 3, and Agent 5 work together to complete the first task.

[0453] The collaborative method provided in this application has been described above with reference to the accompanying drawings. The collaborative device provided in this application will now be described.

[0454] Figure 14 This is a structural schematic diagram of the device provided in an embodiment of this application. Specifically, as shown... Figure 14 As shown, the device 1400 includes: a transceiver module 1401 and a processing module 1402.

[0455] For example, device 1400 can be applied to an ACN-C network element.

[0456] Specifically, the transceiver module 1401 is used to acquire first information, which includes information for instructing the first task; the processing module 1402 is used to: determine N intelligent agents to collaboratively complete the first task based on the first information, where N is a positive integer greater than 1; the transceiver module 1401 is also used to: send first behavior information to the first intelligent agent, which is included among the N intelligent agents, and the first behavior information is used to instruct the behavior of the first intelligent agent when collaborating with other intelligent agents among the N intelligent agents to complete the first task.

[0457] In one implementation, the transceiver module 1401 is further configured to: receive first capability information from the first intelligent agent, the first capability information being used to indicate the capabilities of the first intelligent agent, the capabilities including ontology capabilities and / or model capabilities.

[0458] In one implementation, the transceiver module 1401 is specifically used to: receive first information from the first intelligent agent.

[0459] For example, the device 1400 can be applied to a first intelligent agent.

[0460] Specifically, the transceiver module 1401 is used to receive first action information, which is used to instruct the behavior of the first intelligent agent when collaborating with other intelligent agents among N intelligent agents to complete the first task, where N is a positive integer greater than 1; the processing module 1402 is used to: collaborate with other intelligent agents among N intelligent agents to complete the first task according to the first action information.

[0461] In one implementation, the transceiver module 1401 is further configured to: receive speaking instruction information from the speaking manager, the speaking instruction information being used to instruct the first intelligent agent to speak.

[0462] In one possible implementation, the transceiver module 1401 is further configured to: send a first conflict indication message, the first conflict indication message being used to indicate that the decision result of the first agent conflicts with the decision result of the second agent among the N agents; and / or, receive a second conflict indication message, the second conflict indication message being used to indicate that the decision result of the first agent conflicts with the decision result of the third agent among the N agents; and / or, receive a re-decision indication message, the re-decision indication message being used to instruct the first agent to make a new decision.

[0463] In one possible implementation, the transceiver module 1401 is further configured to: receive negotiation summary information, which includes the decision results of each of the N agents, and the decision results of each agent instruct each agent to perform actions to complete the first task.

[0464] In one possible implementation, the transceiver module 1401 is further configured to: send first information, the first information including information for instructing a first task.

[0465] In one implementation, the transceiver module 1401 is further configured to: send first capability information, the first capability information being used to indicate the capabilities of the first intelligent agent, the capabilities including ontology capabilities and / or model capabilities.

[0466] In one implementation, the first behavioral information is used to instruct the behavior of the first intelligent agent when collaborating with other intelligent agents among N intelligent agents to complete a first task, including: the first behavioral information is used to instruct at least one of the following: the collaboration method when the first intelligent agent collaborates with other intelligent agents to complete the first task, the role undertaken by the first intelligent agent, the message template when the first intelligent agent communicates with other intelligent agents, the identifier of other intelligent agents, and the capabilities of other intelligent agents.

[0467] In one implementation, the first behavioral information is used to indicate the collaboration mode when the first intelligent agent collaborates with other intelligent agents to complete the first task, including: the first behavioral information is used to indicate that the collaboration mode when the first intelligent agent collaborates with other intelligent agents to complete the first task is a centralized collaboration mode.

[0468] In one implementation, the first action information is used to indicate the role assumed by the first intelligent agent, including: the first action information includes instruction information for instructing the first intelligent agent to assume the role of a leader; or, the first action information includes instruction information for instructing the first intelligent agent to assume the role of a member; or, the first action information includes leader recommendation instruction information, which instructs the first intelligent agent to determine the role assumed by the first intelligent agent based on the leader recommendation process.

[0469] In one implementation, if the first action information includes recommended leader instruction information, the first action information also includes a first winning rule; wherein, the first intelligent agent determines the role it assumes based on the first winning rule.

[0470] In one implementation, the first action information is used to indicate the collaboration mode when the first intelligent agent collaborates with other intelligent agents to complete the first task, including: the first action information is used to indicate that the collaboration mode when the first intelligent agent collaborates with other intelligent agents to complete the first task is a distributed collaboration mode.

[0471] In one implementation, the first line of information is further used to indicate at least one of the following: the speaking order of the N agents, the arbitration method when at least two of the N agents have a decision conflict, the data mode used when the N agents negotiate, and the maximum token length used when the N agents negotiate.

[0472] In one implementation, the first line of information is used to indicate the speaking order of N agents, including: the first line of information is used to indicate speaking in a fixed speaking order; or, the first line of information is used to indicate speaking according to the scheduling of the speaking manager.

[0473] In one implementation, the first behavioral information is used to indicate the arbitration method when at least two of the N agents have a decision conflict, including: the first behavioral information is used to indicate arbitration based on the priority of the at least two agents; or, the first behavioral information is used to indicate arbitration based on the arbitration manager.

[0474] In one implementation, the first behavioral information is used to instruct a first intelligent agent to determine the role it assumes, including: the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a member; and / or, if the first behavioral information is used to instruct speaking according to the scheduling of a speaking manager, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of a speaking manager; and / or, if the first behavioral information is used to instruct arbitration based on an arbitration manager, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of an arbitration manager.

[0475] In one implementation, the first action information is used to instruct a first intelligent agent to determine its role, including: if the first action information instructs the agent to speak according to the scheduling of a speech manager, the first action information includes a recommended speech manager instruction, which instructs the first intelligent agent to determine its role based on the recommended speech manager process; and / or, if the first action information instructs the agent to arbitrate based on an arbitration manager, the first action information includes a recommended arbitration manager instruction, which instructs the first intelligent agent to determine its role based on the recommended arbitration manager process.

[0476] In one implementation, the first information further includes at least one of the following: an identifier of a candidate intelligent agent that the first intelligent agent expects to cooperate in completing the first task; and indication information for indicating the cooperative method expected by the first intelligent agent.

[0477] In one implementation, the first intelligent agent expects a centralized collaborative approach, and the first information also includes instruction information indicating that the first intelligent agent expects to assume a leader role.

[0478] In one possible implementation, the method further includes: sending first capability information to the ACN-C network element, the first capability information being used to indicate the capabilities of the first intelligent agent, the capabilities including ontological capabilities and / or model capabilities.

[0479] Figure 15 This is a structural schematic diagram of another device provided in an embodiment of this application. Figure 15 The apparatus shown can be used to perform the methods described in the foregoing embodiments.

[0480] like Figure 15 As shown, the device 1500 of this embodiment includes a memory 1501 and a processor 1502. In one implementation, the device 1500 further includes a communication interface 1503 and a bus 1504. The memory 1501, processor 1502, and communication interface 1503 are interconnected via the bus 1504.

[0481] The memory 1501 may be a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1501 may store programs, and when the program stored in the memory 1501 is executed by the processor 1502, the processor 1502 performs the various steps of the methods described above.

[0482] The processor 1502 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, used to execute relevant programs to implement the various steps of the communication method described above.

[0483] The processor 1502 can also be an integrated circuit chip with signal processing capabilities. In the implementation process, the embodiments of this application... Figures 3 to 13 Each step of the method can be accomplished by integrated logic circuitry in the hardware of the processor 1502 or by instructions in software form.

[0484] The processor 1502 described above can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or a conventional processor, etc.

[0485] The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory 1501. Processor 1502 reads information from memory 1501 and, in conjunction with its hardware, completes the functions required by the units included in the device of this application. For example, it can execute the various steps / functions of the communication method described above.

[0486] The communication interface 1503 can use, but is not limited to, transceivers to enable communication between the device 1500 and other devices or communication networks.

[0487] Bus 1504 may include a pathway for transmitting information between various components of device 1500 (e.g., memory 1501, processor 1502, communication interface 1503).

[0488] It should be understood that the apparatus 1500 shown in the embodiments of this application can be deployed in access network equipment or terminals.

[0489] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be a usable medium accessible to a computer or a data storage device such as a server or data center containing one or more sets of usable media. The usable medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0490] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0491] In this application, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0492] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not limit the implementation process of the embodiments of this application.

[0493] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0494] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0495] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0496] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0497] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0498] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

Claims

1. A method for cooperation among intelligent agents, characterized in that, include: Obtain first information, which includes information for instructing a first task; Based on the first information, N intelligent agents are determined to collaboratively complete the first task, where N is a positive integer greater than 1; Send first action information to the first intelligent agent, which is included among the N intelligent agents. The first action information is used to instruct the behavior of the first intelligent agent when it collaborates with other intelligent agents among the N intelligent agents to complete the first task.

2. The method according to claim 1, characterized in that, The first behavioral information is used to instruct the behavior of the first agent when it collaborates with other agents among the N agents (excluding the first agent) to complete the first task, including: The first behavioral information is used to indicate at least one of the following: the collaboration method when the first intelligent agent collaborates with the other intelligent agents to complete the first task, the role undertaken by the first intelligent agent, the message template when the first intelligent agent communicates with the other intelligent agents, the identifier of the other intelligent agent, and the capabilities of the other intelligent agents.

3. The method according to claim 2, characterized in that, The first behavioral information is used to indicate the collaborative method when the first agent collaborates with the other agents to complete the first task, including: The first behavioral information is used to indicate that the collaboration mode when the first intelligent agent and the other intelligent agents cooperate to complete the first task is a centralized collaboration mode.

4. The method according to claim 3, characterized in that, The first behavioral information is used to indicate the role undertaken by the first intelligent agent, including: The first behavioral information includes instructions to instruct the first intelligent agent to assume a leader role; or, The first behavioral information includes instruction information for instructing the first intelligent agent to assume a member role; or, The first behavioral information includes recommended leader instruction information, which is used to instruct the first agent to determine the role it assumes based on the recommended leader process.

5. The method according to claim 4, characterized in that, If the first behavioral information includes recommended leader instruction information, the first behavioral information also includes a first winning rule; The first intelligent agent determines its role based on the first winning rule.

6. The method according to claim 2, characterized in that, The first behavioral information is used to indicate the collaborative method when the first agent collaborates with the other agents to complete the first task, including: The first behavioral information is used to indicate that the first intelligent agent and the other intelligent agents cooperate in completing the first task in a distributed cooperative manner.

7. The method according to claim 6, characterized in that, The first behavioral information is also used to indicate at least one of the following: the speaking order of the N agents, the arbitration method when at least two of the N agents have a decision conflict, the data mode used by the N agents during negotiation, and the maximum token length used by the N agents during negotiation.

8. The method according to claim 7, characterized in that, The first behavioral information is used to indicate the speaking order of the N agents, including: The first behavioral information is used to indicate that speaking should be done in a fixed order; or... The first behavioral information is used to instruct the speaker to speak according to the speaker manager's schedule.

9. The method according to claim 7 or 8, characterized in that, The first behavioral information is used to indicate the arbitration method when at least two of the N agents have a decision conflict, including: The first behavioral information is used to instruct arbitration based on the priorities of the at least two agents; or, The first behavioral information is used to instruct arbitration to be conducted based on the arbitration manager.

10. The method according to any one of claims 7 to 9, characterized in that, The first behavioral information is used to instruct the first intelligent agent to determine the role it assumes, including: The first behavioral information includes instruction information for instructing the first intelligent agent to assume a member role; and / or, If the first behavioral information is used to instruct the speaker to speak according to the speaker manager's schedule, the first behavioral information includes instruction information to instruct the first agent to assume the role of speaker manager; and / or, If the first behavioral information is used to instruct arbitration to be conducted based on an arbitration manager, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of an arbitration manager.

11. The method according to any one of claims 7 to 9, characterized in that, The first behavioral information is used to instruct the first intelligent agent to determine the role it assumes, including: If the first behavioral information is used to instruct speaking according to the scheduling of the speaking manager, the first behavioral information includes a recommended speaking manager instruction, which instructs the first agent to determine the role it assumes based on the process of recommending a speaking manager; and / or, If the first action information indicates that arbitration is conducted based on an arbitration manager, the first action information includes a recommendation for an arbitration manager instruction, which is used to instruct the first agent to determine the role it assumes based on the process of recommending an arbitration manager.

12. The method according to any one of claims 1 to 11, characterized in that, The method further includes: Receive first capability information from the first intelligent agent, the first capability information being used to indicate the capabilities of the first intelligent agent, the capabilities including ontology capabilities and / or model capabilities.

13. A method for cooperation among intelligent agents, characterized in that, Applied to a first intelligent agent, the method includes: Receive first behavior information, which is used to instruct the behavior of the first agent when it collaborates with other agents among N agents (excluding the first agent) to complete the first task, where N is a positive integer greater than 1; Based on the first behavioral information, the agent collaborates with other agents among the N agents to complete the first task.

14. The method according to claim 13, characterized in that, The first behavioral information is used to instruct the behavior of the first agent when it collaborates with other agents among the N agents (excluding the first agent) to complete the first task, including: The first behavioral information is used to indicate at least one of the following: the collaboration method when the first intelligent agent collaborates with the other intelligent agents to complete the first task, the role undertaken by the first intelligent agent, the message template when the first intelligent agent communicates with the other intelligent agents, the identifier of the other intelligent agent, and the capabilities of the other intelligent agents.

15. The method according to claim 14, characterized in that, The first behavioral information is used to indicate the collaborative method when the first agent collaborates with the other agents to complete the first task, including: The first behavioral information is used to indicate that the collaboration mode when the first intelligent agent and the other intelligent agents cooperate to complete the first task is a centralized collaboration mode.

16. The method according to claim 15, characterized in that, The first behavioral information is used to indicate the role undertaken by the first intelligent agent, including: The first behavioral information includes instructions to instruct the first intelligent agent to assume a leader role; or, The first behavioral information includes instruction information for instructing the first intelligent agent to assume a member role; or, The first behavioral information includes recommended leader instruction information, which is used to instruct the first agent to determine the role it assumes based on the recommended leader process.

17. The method according to claim 15, characterized in that, If the first behavioral information includes recommended leader instruction information, the first behavioral information also includes a first winning rule; The first intelligent agent determines its role based on the first winning rule.

18. The method according to claim 14, characterized in that, The first behavioral information is used to indicate the collaborative method when the first agent collaborates with the other agents to complete the first task, including: The first behavioral information is used to indicate that the first intelligent agent and the other intelligent agents cooperate in completing the first task in a distributed cooperative manner.

19. The method according to claim 18, characterized in that, The first behavioral information is also used to indicate at least one of the following: the speaking order of the N agents, the arbitration method when at least two of the N agents have a decision conflict, the data mode used by the N agents during negotiation, and the maximum token length used by the N agents during negotiation.

20. The method according to claim 19, characterized in that, The first behavioral information is used to indicate the speaking order of the N agents, including: The first behavioral information is used to indicate that speaking should be done in a fixed order; or... The first behavioral information is used to instruct the speaker to speak according to the speaker manager's schedule.

21. The method according to claim 19 or 20, characterized in that, The first behavioral information is used to indicate the arbitration method when at least two of the N agents have a decision conflict, including: The first behavioral information is used to instruct arbitration based on the priorities of the at least two agents; or, The first behavioral information is used to instruct arbitration to be conducted based on the arbitration manager.

22. The method according to any one of claims 19 to 21, characterized in that, The first behavioral information is used to instruct the first intelligent agent to determine the role it assumes, including: The first behavioral information includes instruction information for instructing the first intelligent agent to assume a member role; and / or, If the first behavioral information is used to instruct the speaker to speak according to the speaker manager's schedule, the first behavioral information includes instruction information to instruct the first agent to assume the role of speaker manager; and / or, If the first behavioral information is used to instruct arbitration to be conducted based on an arbitration manager, the first behavioral information includes instruction information for instructing the first intelligent agent to assume the role of an arbitration manager.

23. The method according to any one of claims 19 to 21, characterized in that, The first behavioral information is used to instruct the first intelligent agent to determine the role it assumes, including: If the first behavioral information is used to instruct speaking according to the scheduling of the speaking manager, the first behavioral information includes a recommended speaking manager instruction, which instructs the first agent to determine the role it assumes based on the process of recommending a speaking manager; and / or, If the first action information indicates that arbitration is conducted based on an arbitration manager, the first action information includes a recommendation for an arbitration manager instruction, which is used to instruct the first agent to determine the role it assumes based on the process of recommending an arbitration manager.

24. The method according to any one of claims 19 to 23, characterized in that, If the first behavioral information is used to instruct speaking according to the speaking manager's schedule, the step of collaborating with other agents among the N agents to complete the first task includes: The system receives speaking instruction information from the speaking manager, which is used to instruct the first intelligent agent to speak.

25. The method according to any one of claims 19 to 24, characterized in that, The step of collaborating with other agents among the N agents to complete the first task includes: Send a first conflict indication message, the first conflict indication message being used to indicate that the decision result of the first agent conflicts with the decision result of the second agent among the N agents; and / or, Receive second conflict indication information, the second conflict indication information being used to indicate a conflict between the decision result of the first agent and the decision result of the third agent among the N agents; and / or, Receive re-decision instruction information, which is used to instruct the first agent to make a new decision.

26. A communication device, characterized in that, It includes a module for performing the method as described in any one of claims 1 to 12; or, it includes a module for performing the method as described in any one of claims 13 to 25.

27. A communication device, characterized in that, include: processor, The processor is configured to execute a computer program and / or, via logic circuitry, cause the communication device to implement the method as described in any one of claims 1 to 12; or to cause the communication device to implement the method as described in any one of claims 13 to 25.

28. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a program or instructions that, when executed, cause the method as claimed in any one of claims 1 to 12 to be implemented; or cause the method as claimed in any one of claims 13 to 25 to be implemented.

29. A computer program product, characterized in that, The computer program product includes a computer program that, when run, causes the method as described in any one of claims 1 to 12 to be implemented; or causes the method as described in any one of claims 13 to 25 to be implemented.

Images