A multi-agent system

Abstract

The application provides a multi-agent system and relates to the technical field of agent task cooperative processing. The application comprises a request execution subsystem and a service monitoring subsystem which are connected through a message center module; the service monitoring subsystem is used for collecting target user behavior state data and environment state data and observing and analyzing the target user behavior state data and the environment state data to obtain observation results and analysis results; the request execution subsystem is used for processing a request message of a target user according to the observation results, the analysis results and context information of the request message to obtain a processing result and feeding back the processing result to the target user through the message center module, wherein the user request execution subsystem is a hierarchical structure of a terminal agent, an intermediate agent and a center agent. The application effectively solves the problems of low task division efficiency, resource waste and insufficient risk identification capability in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent agent task coordination processing, in particular to a multi-agent system. BACKGROUND

[0002] The current multi-agent system generally adopts a flat architecture or a simple hierarchical structure, and has the following significant defects in task processing.

[0003] Extensive task division and low processing efficiency: the existing system lacks a dynamic task allocation mechanism and cannot automatically match intelligent agents according to task complexity (such as simple tasks still requiring central intelligent agents to process), resulting in resource waste.

[0004] Lack of standardized switching mechanism for intelligent agent collaboration, inefficient cross-domain collaboration, and intelligent agent islands: the task switching between intelligent agents in the existing system relies on simple routing rules such as intent classification, and lacks dynamic matching ability based on metadata. When the task exceeds the processing range of the current intelligent agent, it cannot intelligently select the optimal switching object according to the "responsibility matching degree" and "tool set compatibility", often leading to switching failure or repeated switching; when the user proposes a cross-domain or composite demand, due to the lack of intermediate coordination mechanism, the switching is not smooth, resulting in the inability of intelligent agents to fully exert their capabilities.

[0005] Insufficient abnormality monitoring and risk identification capability: the traditional system does not set up an independent supervision system, and lacks real-time perception and self-adaptive adjustment capability for changes in user and environment state data, process tracking (such as user request overdue without solution), etc. For example, when the terminal intelligent agent returns error data, the system cannot timely trigger the fuse mechanism or dynamically switch to the alternative solution, which easily leads to task interruption or user experience degradation.

[0006] Context missing: the existing message transmission mechanism does not establish the association relationship of user-session-request, and cannot effectively track the complete context information of user request during cross-task collaboration. SUMMARY

[0007] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a multi-agent system, which solves the problems of low task division processing efficiency, resource waste and insufficient risk identification capability in the prior art.

[0008] To achieve the above purpose, the present application provides the following scheme:

[0009] A multi-agent system, comprising:

[0010] A request execution subsystem and a service monitoring subsystem connected in communication by a message center module;

[0011] The service monitoring subsystem is configured to collect target user behavior state data and environment state data, and to observe and analyze the target user behavior state data and the environment state data to obtain observation results and analysis results. The request execution subsystem is configured to process a request message of a target user according to the observation results, the analysis results, and context information of the request message, to obtain a processing result, and to feed back the processing result to the target user through the message center module. The user request execution subsystem is a hierarchical structure of a terminal agent, an intermediate agent, and a center agent.

[0012] Preferably, the state of the request message of the target user includes:

[0013] User request overdue, terminal agent returning error data, normal user request processing, and abnormal user request processing.

[0014] Preferably, the request message of the target user is multi-modal data.

[0015] Preferably, the multi-modal data includes:

[0016] Text data, voice data, and image data.

[0017] Preferably, the service monitoring subsystem includes:

[0018] An observation agent and a supervision agent.

[0019] The observation agent is connected to the supervision agent through a message center module.

[0020] The observation agent is configured to collect behavior state data and environment state data of the target user, and the supervision agent is configured to observe and analyze the behavior state data and the environment state data of the target user to obtain observation results and analysis results.

[0021] Preferably, the request execution subsystem includes:

[0022] A plurality of terminal agents, a plurality of intermediate agents, and a center agent, which are all communicatively connected through a message center module.

[0023] The terminal agent is used for processing the request message of the target user in combination with the observation result and the analysis result, obtaining a first processing sub-result, if the first processing sub-result is processing completed, generating a suggestion according to the first processing sub-result and sending the suggestion to the target user, if the first processing sub-result is not completed, performing transfer judgment, obtaining a judgment result, if the judgment result is needed, identifying a terminal agent capable of processing the request of the target user according to the context of the request message and the metadata of each agent in the transfer address book, if the judgment result is not needed, sending the current task to an intermediate agent, the intermediate agent is used for re-distributing the terminal agent according to the first processing result, obtaining a plurality of second processing sub-results and integrating the second processing sub-results, obtaining an integrated result, if the integrated result is task completed, generating a corresponding suggestion and sending the suggestion to the target user, if the integrated result is not completed, sending the current request message to the center agent, the center agent is used for processing the current request message according to the integrated result, the observation result and the analysis result, obtaining a third processing sub-result and generating a corresponding suggestion and sending the suggestion to the target user.

[0024] The present application discloses the following technical effects:

[0025] The present application provides a multi-agent system, comprising:

[0026] The request execution subsystem and the service monitoring subsystem are communicatively connected through the message center module;

[0027] The service monitoring subsystem is used for collecting target user behavior state data and environment state data and observing and analyzing the target user behavior state data and the environment state data, obtaining an observation result and an analysis result, the request execution subsystem is used for processing the request message of the target user according to the observation result, the analysis result and the context information of the request message, obtaining a processing result and feeding back the processing result to the target user through the message center module, wherein the user request execution subsystem is a hierarchical structure of a terminal agent, an intermediate agent and a center agent. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments will be briefly introduced. Obviously, the accompanying drawings in the following description only only some embodiments of the present application, and all other embodiments obtained by a person of ordinary skill in the art without creative work based on the accompanying drawings belong to the protection scope of the present application.

[0029] Figure 1 A multi-agent data flow diagram provided for the embodiments of the present application;

[0030] Figure 2 An agent relationship diagram provided for the embodiments of the present application;

[0031] Figure 3 A message structure diagram provided for the embodiments of the present application;

[0032] Figure 4 A data flow diagram of a terminal agent directly processing a simple task provided for the embodiments of the present application;

[0033] Figure 5 A flowchart of an agent processing a message provided for the embodiments of the present application;

[0034] Figure 6 A government affair agent system relationship diagram provided for the embodiments of the present application. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work belong to the protection scope of the present application.

[0036] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail with reference to the accompanying drawings and specific embodiments.

[0037] As shown in Figures 1-5 , the present application provides a multi-agent system, comprising:

[0038] A request execution subsystem and a service monitoring subsystem connected in communication through a message center module;

[0039] The service monitoring subsystem is configured to collect target user behavior state data and environment state data, and to observe and analyze the target user behavior state data and the environment state data to obtain observation results and analysis results. The request execution subsystem is configured to process a request message of a target user according to the observation results, the analysis results, and context information of the request message, to obtain a processing result, and to feed back the processing result to the target user through the message center module. The user request execution subsystem is a hierarchical structure of a terminal intelligent agent, an intermediate intelligent agent, and a center intelligent agent.

[0040] Specifically, identifying the operation behavior of the user and the environmental parameters includes: using a video image recognition model to identify whether there is a human face in front of the camera, and identifying the expression of the human face; identifying the expression according to the voice through a voice model; and identifying the passenger flow in the surrounding environment.

[0041] After identifying the operation behavior of the user and the environmental parameters, the observation intelligent agent saves the observation results, the parallel user, and the time stamp to an observation record database.

[0042] Further, the state of the request message of the target user includes:

[0043] User request overdue, terminal intelligent agent returns error data, user request processing is normal, and user request processing is abnormal.

[0044] Further, the request message of the target user is multi-modal data.

[0045] The multi-modal data includes:

[0046] Text data, voice data, and image data.

[0047] Specifically, the multi-modal data is generally text, voice, and image,

[0048] Uncommon multi-modal data also includes olfaction and taste.

[0049] Other integrated sensors can achieve multi-modal data, including temperature, temperature, and air pressure.

[0050] Other multi-modal data obtained through further processing based on images includes user emotion, identity, gender, age, action, expression, passenger flow, and surrounding environment.

[0051] Other multi-modal data obtained through further processing based on audio includes noise level and voiceprint.

[0052] Further, the service monitoring subsystem includes:

[0053] An observation intelligent agent and a supervision intelligent agent.

[0054] The observation agent is connected with the supervision agent through a message center module;

[0055] The observation agent is used to collect the behavior state data and the environment state data of the target user, and the supervision agent is used to observe and analyze the behavior state data and the environment state data of the target user to obtain observation results and analysis results.

[0056] Further, the request execution subsystem comprises:

[0057] A plurality of terminal agents, a plurality of intermediate agents and a center agent are connected through a message center module;

[0058] The terminal agent is used to process the request message of the target user in combination with the observation results and the analysis results to obtain a first processing sub-result, if the first processing sub-result is completed, a suggestion is generated according to the first processing sub-result and sent to the target user, if the first processing sub-result is not completed, a transfer judgment is performed to obtain a judgment result, if the judgment result is needed, the terminal agent capable of processing the request of the target user is identified according to the context of the request message and the metadata of each agent in the transfer address book, if the judgment result is not needed, the current task is sent to the intermediate agent, the intermediate agent is used to re-allocate the terminal agent according to the first processing result to process the current request message, obtain a plurality of second processing sub-results and integrate the second processing sub-results to obtain an integrated result, if the integrated result is completed, a corresponding suggestion is generated and sent to the target user, if the integrated result is not completed, the current request message is sent to the center agent, the center agent is used to process the current request message according to the integrated result, the observation result and the analysis result to obtain a third processing sub-result and generate a corresponding suggestion and send it to the target user.

[0059] Specifically, the multi-agent system processes user requests through dialogue. The initiation message of a user's request usually includes a text or voice question and accompanying multi-modal information such as documents and images. A session (request information) between a user and the multi-agent system includes multiple requests. A request includes a message sent by the user to the terminal agent, zero to multiple messages between agents, and one to multiple messages sent by the agent to the user. Messages between the user and the agent, messages between the agent and the user, and messages between agents are sent and received through a message center. Agent metadata includes: responsibilities, a set of callable tools, a superior agent, and a directory of switchable agents. The service monitoring subsystem collects and records user behavior state data and its environment state data by observation agents. The analysis results are sent to the user or other agents by the supervisory agent through the message center. The terminal agent, intermediate agent, and central agent receive messages while obtaining observation records related to the messages. The messages and their context and observation records are processed. The terminal agent and intermediate agent process messages according to their responsibilities and callable tool sets. If the agent can handle the message, the processing result is sent to the user through the message center. If the agent cannot handle the message, a switch message is sent to other agents or a superior agent through the message center.

[0060] Further, the agent is described by its metadata, which includes the following dimensions: responsibilities, callable tools, superiors, and switch directories.

[0061] Responsibilities: Clearly define the core function range and task objectives of the agent. For example, the responsibility of the terminal agent is to handle simple user inquiries, and the responsibility of the intermediate agent is to coordinate cross-domain tasks. Clear responsibility definitions ensure that tasks can be accurately assigned to appropriate agents.

[0062] Tool set: Tools that the agent can call to complete tasks, such as MCP services. Different types of agents have different tool sets to meet their needs for handling specific tasks.

[0063] Superior: The immediate superior of the agent, a higher-level agent, which clarifies the hierarchical relationship of the agent, facilitating task uploading and assignment.

[0064] Switch directory: A list of switchable agents and their applicable scenarios when a task exceeds the current agent's capabilities or needs to be assigned to other agents. For example, when the terminal agent encounters a complex task, it can select an appropriate intermediate agent for switching according to the switch directory.

[0065] Explanation of multi-agent data transmission in the accompanying drawings:

[0066] Figure 3n represents the number of messages.

[0067] Ma1: User submits a request message, containing user question, receiver, etc. This is the starting point of the task, and the user puts forward the demand to the system through the message.

[0068] Ma2: Terminal agent receives data, including messages sent to itself and its context, and observation records of the initiating user. The terminal agent obtains complete information related to the task, and prepares for processing the task.

[0069] Ma3: Terminal agent sends messages, containing processing results, receivers, etc. If the terminal agent can process the task, it sends the processing results to the corresponding receivers, such as users or superior agents.

[0070] Ma4: Intermediate agent receives data, including messages sent to itself and its context, and observation records of the initiating user. The intermediate agent obtains information related to cross-domain tasks, and coordinates and processes them.

[0071] Ma5: Intermediate agent sends messages, containing processing results, receivers, etc. The intermediate agent sends the processed results to related agents or users.

[0072] Ma6: Central agent receives data, including messages sent to itself and its context, and observation records of the initiating user. The central agent obtains global information, and makes important decisions and strategies.

[0073] Ma7: Central agent sends messages, containing processing results, receivers, etc. The central agent sends decisions and strategies to related agents, guiding the operation of the system.

[0074] Ma8: User receives messages sent to itself, and obtains task processing results.

[0075] Mb1: Observation agent sends observation records of the user, providing user state information for supervisory agents and central agents.

[0076] Mb2: Supervisory agent analyzes messages and observation records, identifying abnormalities and risks.

[0077] Mb3: Supervisory agent sends messages to specific receivers, such as users requesting to handle exceptions, sending notifications to related agents or users.

[0078] Further, the request message includes:

[0079] User ID, session ID, request ID, sender ID, receiver ID, message time, and message body.

[0080] Further, the message metadata is described as follows:

[0081] User ID: Identifies the user initiating the conversation, ensuring the system can accurately identify the user's identity.

[0082] Session ID: Identifies a single continuous interaction process, facilitating tracking of all requests and responses by the user in a continuous interaction.

[0083] Request ID: Identifies the user's single request, clearly defining the uniqueness of each request.

[0084] Sender ID: Identifies the message source, which can be a user ID or an agent ID, facilitating the determination of the message's origin.

[0085] Receiver ID: Identifies the message target, which can be a user ID or an agent ID, ensuring accurate message delivery.

[0086] Message Time: Message sending time, recording the timestamp of the message for time series analysis and abnormal tracking.

[0087] Message Body: Message content, including specific task information, processing results, etc.

[0088] A user can initiate one or more sessions, a session contains one or more requests, and a request consists of one or more messages. This hierarchical relationship establishes the complete context of user requests, ensuring effective tracking of complete information of user requests when cross-task collaboration.

[0089] Further, the process of each agent processing messages is as follows:

[0090] s1: Receive message: The agent receives messages from users or other agents.

[0091] s2: Process message: The agent processes the message according to its own responsibilities and tool set.

[0092] s3: Determine complete: Determine whether the processing is complete.

[0093] Yes → s4: If the processing is complete, send a response message to the user, informing the user of the processing result.

[0094] No → s5: If the processing is not complete, enter the judgment transfer link.

[0095] s5: Determine transfer: Determine whether it needs to be transferred to other agents.

[0096] Yes → s6: If it needs to be transferred, send a transfer message to other agents according to the transfer directory, and select the optimal transfer object.

[0097] (NO) -> s7: If no forwarding is needed, go to decision transfer superior.

[0098] s7: Decision transfer superior: Determine whether to transfer to a superior agent.

[0099] (YES) -> s8: If transfer to a superior is needed, send a transfer superior message to the superior agent.

[0100] (NO) -> s9: If no transfer to a superior is needed, generate a suggestion.

[0101] s9: Generate suggestion: Generate a suggestion for the user based on the processing.

[0102] s10: Send suggestion message to user: Send the suggestion to the user to provide further guidance to the user.

[0103] Further, as shown in Figure 6 , the present application also provides another embodiment of a government affair agent system, wherein the terminal agent is a human resource digital person large screen, and the model of the built-in model is qwen 7B; the intermediate agent includes a human resource expert agent and a policy expert agent; wherein the model of the built-in model of the intermediate agent is qwen 32B. The observation agent is a digital person large screen vision module, wherein the built-in model of the observation agent is janus 7B. The center agent is a government affair agent, wherein the model of the built-in model of the center agent is deepseek 671B.

[0104] In the specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between each embodiment can be referred to each other.

[0105] In this paper, specific examples are used to illustrate the principles and implementation methods of the present application. The above description of the embodiments is only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation method and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A multi-agent system, characterized in that, The application relates to a message center module, a request execution subsystem and a service monitoring subsystem. The service monitoring subsystem is used for collecting and observing and analyzing the behavior state and the environment state of a target user, obtaining observation results and analysis results, the request execution subsystem is used for processing a request message of the target user according to the observation results, the analysis results and context information of the request message, obtaining a processing result and feeding back the processing result to the target user through the message center module, and the request execution subsystem is a hierarchical structure of a terminal intelligent agent, an intermediate intelligent agent and a center intelligent agent. The service monitoring subsystem comprises: An observation intelligent agent and a supervision intelligent agent. The observation intelligent agent is connected with the supervision intelligent agent through the message center module. The observation intelligent agent is used for collecting the behavior state and the environment state of the target user, and the supervision intelligent agent is used for observing and analyzing the behavior state and the environment state of the target user, obtaining observation results and analysis results. The request execution subsystem comprises: A plurality of terminal intelligent agents, a plurality of intermediate intelligent agents and a center intelligent agent which are all connected through the message center module. The terminal intelligent agent is used for processing a request message of a target user in combination with observation results and analysis results, obtaining a first processing subresult, generating a suggestion according to the first processing subresult and sending the suggestion to the target user if the first processing subresult is completed, performing transfer judgment to obtain a judgment result if the first processing subresult is not completed, identifying a terminal intelligent agent capable of processing the request of the target user according to context information of the request message and metadata of intelligent agents in a transfer address book if the judgment result is needed, sending a current task to an intermediate intelligent agent if the judgment result is not needed, the intermediate intelligent agent is used for reassigning terminal intelligent agents according to the first processing subresult, obtaining a plurality of second processing subresults and integrating the second processing subresults to obtain an integrated result, generating a corresponding suggestion and sending the suggestion to the target user if the integrated result is completed, and sending a current request message to the center intelligent agent if the integrated result is not completed, the center intelligent agent is used for processing a current request message according to the integrated result, the observation results and the analysis results, obtaining a third processing subresult and generating a corresponding suggestion and sending the suggestion to the target user. The state of the request message of the target user comprises:

2. The multi-agent system of claim 1, wherein, User request overdue, terminal intelligent agent returning error data, user request processing normally and user request processing abnormally. The request message of the target user is multi-modal data.

3. The multi-agent system of claim 1, wherein, The multi-modal data comprises:

4. The multi-agent system of claim 3, wherein, Text data, voice data and image data. ​

Images