Multi-agent cooperative task planning method, system and device and storage medium

Abstract

The invention provides a multi-agent cooperative task planning method, system and device and a storage medium, and the method comprises the steps: constructing an asynchronous communication framework, responding to a task event through an event-driven model, and decoupling the production and consumption of the task event through a message buffering queue; performing condition matching on the task event based on a predefined behavior rule, if matching succeeds, commanding the intelligent agent to execute an action corresponding to the rule, otherwise, marking the task event as an unmatched task, and transmitting the unmatched task to an LLM language model; loading the historical process data of the unmatched task into the historical message context of the LLM language model, generating a task allocation decision in combination with the guide template, and optimizing the task allocation decision through an abnormal re-decision mechanism to obtain a decision result; and distributing a decision result to a corresponding agent for execution, and updating context data of the task event and a predefined behavior rule in real time. The method can significantly improve the reliability, efficiency and flexibility of task execution, and is suitable for various complex business scenes.

Description

Technical Field

[0001] The present application relates to the field of multi-agent collaborative technology, and specifically to a multi-agent collaborative task planning method, system, device and storage medium. Background Art

[0002] In the field of collaborative task management of multi-agent systems, with the increase in task complexity and the number of agents, traditional architecture mechanisms have gradually exposed many problems. In the early centralized collaborative architecture, all agents communicated and assigned tasks through a central controller, which worked in small-scale scenarios. However, when the number of agents increased, the performance bottleneck of the central controller became prominent, the task processing efficiency was low, and there was a risk of single point failure. Once the central controller failed, the entire system might be paralyzed.

[0003] Although the direct collaborative architecture avoids the bottleneck problem of the central controller, the direct communication between intelligent agents lacks a global perspective, making it difficult to carry out effective overall planning. It is easy to fall into local optimality and task conflicts occur frequently, resulting in low task execution efficiency and difficulty in coping with complex and changing task scenarios.

[0004] With the development of artificial intelligence technology, large language models (LLMs) have shown great potential in natural language processing and intelligent decision-making. Introducing LLMs into multi-agent collaborative task planning and combining them with rule engines is expected to overcome the shortcomings of traditional architectures. However, how to efficiently integrate LLMs with rule engines to build a stable, efficient, and flexible multi-agent collaborative task planning system is still an urgent problem to be solved.

[0005] In view of this, the present invention proposes a multi-agent collaborative task planning method, system, device and storage medium, which can provide a better solution for multi-agent collaborative task planning in complex business scenarios. Summary of the invention

[0006] In order to solve the problems of local optimality or conflict that may easily occur in direct communication and collaboration between intelligent agents in complex scenarios, the present invention provides a multi-agent collaborative task planning method, system, device and storage medium to solve the above-mentioned technical defects.

[0007] In a first aspect, the present invention proposes a multi-agent collaborative task planning method, the method comprising the following steps: S1. Build an asynchronous communication framework, respond to task events through an event-driven model, use message buffer queues to decouple task event production and consumption, and realize asynchronous communication between the central controller and multiple agents; S2, conditionally match the task event based on the predefined behavior rules. If the match is successful, the agent is commanded to execute the action corresponding to the rule. Otherwise, the task event is marked as an unmatched task and the unmatched task is transmitted to the LLM language model. S3, loading the historical process data of the unmatched tasks into the historical message context of the LLM language model, generating task allocation decisions in combination with the guidance template, and optimizing the task allocation decisions through the abnormal re-decision mechanism to obtain the decision results; S4. Distribute the decision results to the corresponding intelligent agents for execution, and update the context data and predefined behavior rules of the task events in real time.

[0008] Preferably, in step S1, the task event is responded to through an event-driven model, wherein the event-driven model includes: an event producer, an event distributor, and an event consumer; The event producer generates task events and passes them to the event distributor; The event distributor distributes the task event to the corresponding agent based on the type of the task event; Event consumers call the rule library to match conditions or call the LLM language model to make task decisions, and feed back the processing results to the context data of the task event.

[0009] Preferably, task events include task initiation, exception triggering, decision completion, and multi-agent collaborative messages.

[0010] Preferably, in step S3, the historical process data of the unmatched tasks is loaded into the historical message context of the LLM language model, and a task allocation decision is generated in combination with the guidance template, which specifically includes the following sub-steps: S31, dynamically load the LLM language model configuration, and load the historical process data of the unmatched tasks into the historical message context of the LLM language model; S32. Calling a preset guidance model according to the task type of the unmatched task to generate a task allocation decision, wherein the guidance template includes a description of the duties of the agent members and structured instructions for task allocation.

[0011] Preferably, in step S3, the task allocation decision is optimized through an abnormal re-decision mechanism to obtain a decision result, including: Determine whether the task allocation decision is an abnormal situation. If it is, trigger the abnormal re-decision mechanism and optimize it through the following expression:

[0012] In the formula, represents the optimized decision result, represents the correction context under abnormal conditions; H represents the task history process data, which serves as context input; Represents a boot template optimized for abnormal situations.

[0013] Preferably, step S1 further includes: using an asyncio library to create a coroutine task, and utilizing a single-threaded event loop to schedule task event execution.

[0014] Preferably, in step S2, conditional matching is performed on the task event based on the predefined behavior rules, and if the match is successful, the agent is commanded to execute the action corresponding to the rule, which also includes: S21, define a rule engine class, and receive predefined behavior rules during initialization; S22, traverse each predefined behavior rule and check whether the task event meets the conditions of the rule; S23. If the task event meets the conditions of the rule, the agent is commanded to execute the action corresponding to the rule and the processing of this task event is ended.

[0015] In a second aspect, the present invention proposes a multi-agent collaborative task planning system, the system comprising: The asynchronous communication interface module is configured to build an asynchronous communication framework, respond to task events through an event-driven model, use the message buffer queue to decouple task event production and consumption, and realize asynchronous communication between the central controller and multiple agents; The rule engine module matches the task events based on predefined behavior rules. If the match is successful, the agent is commanded to execute the action corresponding to the rule. Otherwise, the task event is marked as an unmatched task and the unmatched task is transferred to the LLM language model. The LLM decision-making auxiliary module is configured to load the historical process data of unmatched tasks into the historical message context of the LLM language model, generate task allocation decisions in combination with the guidance template, and optimize the task allocation decisions through the abnormal re-decision mechanism to obtain the decision results; The agent execution module is configured to distribute the decision results to the corresponding agents for execution and update the context data and predefined behavior rules of the task events in real time.

[0016] In the third aspect, the present invention proposes a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multi-agent collaborative task planning method as described in any one of the above items when executing the computer program.

[0017] In a fourth aspect, the present invention proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the multi-agent collaborative task planning method as described in any one of the above items.

[0018] Compared with the prior art, the beneficial results of the present invention are: (1) Solving the defects of traditional architecture: The present invention constructs an asynchronous communication framework, combines event-driven models and message buffer queues, and effectively solves the performance bottleneck and single-point failure risk problems of the central controller in the traditional centralized collaborative architecture, as well as the local optimal or conflict problems caused by direct communication between agents in the direct collaborative architecture, thereby significantly improving the reliability and efficiency of multi-agent collaborative task planning.

[0019] (2) Efficient task processing and decision-making: Through the collaborative work of predefined behavioral rules and LLM language models, the present invention can quickly process routine tasks, while using the powerful intelligent decision-making capabilities of the LLM language model to handle complex tasks and abnormal situations. This combination not only improves the speed of task processing, but also enhances the accuracy and intelligence level of decision-making.

[0020] (3) Dynamic adaptation and optimization: The present invention introduces an abnormal re-decision mechanism, which can automatically trigger the re-decision process when an abnormal situation is detected. Combined with the corrected context and the optimized guidance template, the task allocation decision is intelligently adjusted and optimized to ensure that the task can be smoothly executed and completed, thereby improving the robustness and adaptability of the system.

[0021] (4) Real-time update and optimization: By updating the context data and predefined behavior rules of task events in real time, the present invention ensures that the intelligent agent can make decisions based on the latest information and rules when performing tasks, supports dynamic adjustment of task execution strategies, and improves the flexibility and efficiency of task execution.

[0022] (5) Flexible expansion and integration: The system architecture proposed in this invention is highly scalable and can easily integrate new agents and task types to adapt to business scenarios of different scales and complexities, thus facilitating future system upgrades and expansions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 is a schematic diagram of a multi-agent asynchronous communication and centralized collaboration architecture according to the present invention; Figure 2 is a flow chart of a multi-agent collaborative task planning method according to the present invention; Figure 3 is a schematic diagram of the interaction between the producer-consumer model and the message queue in the asynchronous communication framework according to the present invention; Figure 4 It is a structural diagram of a computer system suitable for implementing an electronic device of an embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0025] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0026] Figure 1 The schematic diagram of the multi-agent asynchronous communication and centralized collaborative architecture of the present invention is shown. Figure 2 The flowchart of the multi-agent collaborative task planning method of the present invention is shown in combination with reference Figure 1 and Figure 2 The present invention proposes a multi-agent collaborative task planning method, comprising the following steps: S1. Build an asynchronous communication framework, respond to task events through an event-driven model, use message buffer queues to decouple task event production and consumption, and implement asynchronous communication between the central controller and multiple intelligent agents (indicated by Agent in the attached figure, similar expressions are used below; Agent intelligent agents).

[0027] In this embodiment, the event-driven model is used to respond to various task events in the system (such as task start, exception trigger, decision completion, multi-agent collaborative messages, etc.). Its core components include: event producer, event distributor and event consumer.

[0028] Among them, the event producer is responsible for generating task events (such as task requests, abnormal signals, etc.) and passing the task events to the event distributor; The event distributor is used to accept the task time generated by the time producer and distribute the task event to the corresponding agent based on the type of the task event; The event consumer is responsible for processing the task time, calling the rule library for condition matching or calling the LLM language model for task decision-making, and feeding back the processing results to the context data of the task event.

[0029] It should be understood that all intelligent agents and task collaborative planning modules (rule base, LLM language model) can act as generators and consumers of events (messages).

[0030] Figure 3 The following is a schematic diagram showing the interaction between the producer-consumer model and the message queue in the asynchronous communication framework of the present invention. Figure 3As shown in the figure, the message buffer queue is used to decouple event producers and consumers to ensure the stability and reliability of the system in high-concurrency scenarios. Its core features include: Asynchronous communication: Asynchronous communication between event producers and event consumers is achieved through message queues to avoid task blocking.

[0031] Message persistence: Ensure that messages are not lost when the system fails or restarts.

[0032] Priority management: supports sorting messages according to task priority to ensure that high-priority tasks are processed first.

[0033] In step S1, it also includes: creating a coroutine task through asyncio.create_task (a function in the Python asyncio library) to achieve non-blocking task execution. Coroutine is a concurrency mechanism within a single thread. It schedules tasks through the event loop to avoid the overhead of thread switching. The memory usage of each coroutine is usually between a few KB and tens of KB.

[0034] Continue to refer Figure 1 and Figure 2 The multi-agent collaborative task planning method proposed by the present invention further includes the following steps: S2, conditionally match the task event based on the predefined behavior rules. If the match is successful, the agent is commanded to execute the action corresponding to the rule. Otherwise, the task event is marked as an unmatched task and the unmatched task is transferred to the LLM language model. That is, all task events are first planned by the predefined behavior rules, and the tasks that are not planned by the predefined behavior rules are decided by the LLM language model.

[0035] In this embodiment, conditional matching is performed on task events based on predefined behavior rules. If the match is successful, the agent is commanded to execute the action corresponding to the rule, which specifically includes the following sub-steps: S21, define a rule engine class through Python code, and receive predefined behavior rules during initialization; S22, traverse each predefined behavior rule and check whether the task event meets the conditions of the rule; S23. If the task event meets the conditions of the rule, the agent is commanded to execute the action corresponding to the rule (for example, directly assigning the agent to handle routine tasks and forwarding tasks that cannot be decided to the LLM language model) and the processing of this task event is ended.

[0036] S3, loading the historical process data of the unmatched tasks into the historical message context of the LLM language model, generating a task allocation decision in combination with the prompt template, and optimizing the task allocation decision through the abnormal re-decision mechanism (indicated by planning in the figure) to obtain the decision result; S4. Distribute the decision results to the corresponding intelligent agents for execution, and update the context data and predefined behavior rules of the task events in real time.

[0037] In step S3, the historical process data of the unmatched tasks is loaded into the historical message context of the LLM language model, and the task allocation decision is generated in combination with the guidance template, which specifically includes the following sub-steps: S31, dynamically load the LLM language model configuration, and load the historical process data of the unmatched tasks into the historical message context of the LLM language model; S32. Calling a preset guidance model according to the task type of the unmatched task to generate a task allocation decision, wherein the guidance template includes a description of the duties of the agent members and structured instructions for task allocation.

[0038] Optimize task allocation decisions through the abnormal re-decision mechanism to obtain decision results, including: Determine whether the task allocation decision is an abnormal situation. If it is, trigger the abnormal re-decision mechanism and optimize it through the following expression:

[0039] In the formula, represents the optimized decision result, represents the correction context under abnormal conditions; H represents the task history process data, which serves as context input; Represents a boot template optimized for abnormal situations.

[0040] The abnormal re-decision mechanism means that the collaborative module splits the task into multiple subtasks and assigns them to different agents for execution. Each agent will receive the corresponding subtask requirements: <Agent>: <Subtask requirements>.

[0041] The LLM language model evaluates whether the task has achieved the expected goal based on the execution results returned by the agent. If it has not achieved the expected goal or the agent returns an error, the collaborative planning module (indicated by planning in the figure) will understand the intent of the current task status through the LLM language model and re-plan the subtask allocation to ensure the smooth completion of the task.

[0042] Whether it is an abnormal situation is determined by the LLM language model's execution results of the agent subtask.

[0043] In a second aspect, the present invention proposes a multi-agent collaborative task planning system, the system comprising: The asynchronous communication interface module is configured to build an asynchronous communication framework, respond to task events through an event-driven model, use the message buffer queue to decouple task event production and consumption, and realize asynchronous communication between the central controller and multiple agents; The rule engine module matches the task events based on predefined behavior rules. If the match is successful, the agent is commanded to execute the action corresponding to the rule. Otherwise, the task event is marked as an unmatched task and the unmatched task is transferred to the LLM language model. The LLM decision-making auxiliary module is configured to load the historical process data of unmatched tasks into the historical message context of the LLM language model, generate task allocation decisions in combination with the guidance template, and optimize the task allocation decisions through the abnormal re-decision mechanism to obtain the decision results; The agent execution module is configured to distribute the decision results to the corresponding agents for execution and update the context data and predefined behavior rules of the task events in real time.

[0044] The present invention proposes an asynchronous centralized collaboration + LLM decision assistance + predefined behavior rule fusion framework, which combines the intelligent decision-making ability of a large language model with the efficient rule processing ability of a rule engine to achieve collaborative optimization of tasks in complex scenarios. Based on the asyncio coroutine mechanism and message buffer queue, the task processing bottleneck in high-concurrency scenarios is solved. Through the exception handling process, the robustness of the system and the quality of collaborative task completion are significantly improved. Through historical context loading and customized prompt templates, the personalization and precision of collaborative task decisions are achieved. By processing routine tasks through the rule engine, the load of the LLM module is reduced and the overall efficiency of the system is improved.

[0045] In the third aspect, the present invention proposes a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multi-agent collaborative task planning method as described in any one of the above items when executing the computer program.

[0046] In a fourth aspect, the present invention proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the multi-agent collaborative task planning method as described in any one of the above items.

[0047] Reference below Figure 4 , which shows a schematic diagram of the structure of a computer system 400 of a terminal device or server suitable for implementing an embodiment of the present application. Figure 4The terminal device or server shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0048] like Figure 4 As shown, the computer system 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage part 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the computer system 400 are also stored. The CPU 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0049] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed, so that a computer program read therefrom is installed into the storage section 408 as needed.

[0050] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 409, and / or installed from the removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, the above functions defined in the method of the present application are executed. It should be noted that the computer-readable medium described in the present application can be a computer-readable signal medium or a computer-readable medium or any combination of the above two. The computer-readable medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, an apparatus or a device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, an apparatus or a device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0051] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as C or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0052] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0053] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A multi-agent collaborative task planning method, characterized in that: The following steps are involved: S1. Build an asynchronous communication framework, respond to task events through an event-driven model, use message buffer queues to decouple task event production and consumption, and realize asynchronous communication between the central controller and multiple agents; S2, conditionally matching the task event based on predefined behavior rules, and if the match is successful, commanding the agent to execute the action corresponding to the rule, otherwise marking the task event as an unmatched task, and transmitting the unmatched task to the LLM language model; S3, loading the historical process data of the unmatched task into the historical message context of the LLM language model, generating a task allocation decision in combination with a guidance template, and optimizing the task allocation decision through an abnormal re-decision mechanism to obtain a decision result; S4. Distribute the decision result to the corresponding intelligent agent for execution, and update the context data of the task event and the predefined behavior rules in real time.

2. The multi-agent collaborative task planning method according to claim 1, characterized in that: In step S1, responding to a task event through an event-driven model, wherein the event-driven model includes: an event producer, an event distributor, and an event consumer; The event producer generates a task event and transmits the task event to the event distributor; The event distributor distributes the task event to the corresponding agent based on the type of the task event; The event consumer calls the rule library to perform condition matching or calls the LLM language model to perform task decision-making, and feeds back the processing result to the context data of the task event.

3. The multi-agent collaborative task planning method according to claim 2, characterized in that: The task events include task initiation, exception triggering, decision completion and multi-agent collaborative messages.

4. The multi-agent collaborative task planning method according to claim 1, characterized in that: In step S3, the historical process data of the unmatched task is loaded into the historical message context of the LLM language model, and a task allocation decision is generated in combination with a guidance template, which specifically includes the following sub-steps: S31, dynamically loading the LLM language model configuration, and loading the historical process data of the unmatched task into the historical message context of the LLM language model; S32. Call the preset guidance model according to the task type of the unmatched task to generate a task allocation decision, wherein the guidance template includes a description of the duties of the agent members and structured instructions for task allocation.

5. The multi-agent collaborative task planning method according to claim 1, characterized in that: In step S3, the task allocation decision is optimized through an abnormal re-decision mechanism to obtain a decision result, including: Determine whether the task allocation decision is an abnormal situation. If so, trigger the abnormal re-decision mechanism and optimize it through the following expression: In the formula, represents the optimized decision result, represents the correction context under abnormal conditions; H represents the task history process data, which serves as context input; Represents a boot template optimized for abnormal situations.

6. The multi-agent collaborative task planning method according to claim 1, characterized in that: In step S1, it also includes: using the asyncio library to create a coroutine task, and using a single-threaded event loop to schedule the execution of the task event.

7. The multi-agent collaborative task planning method according to claim 1, characterized in that: In step S2, condition matching is performed on the task event based on predefined behavior rules, and if the match is successful, the agent is commanded to execute the action corresponding to the rule, which also includes: S21, define a rule engine class, and receive predefined behavior rules during initialization; S22, traversing each predefined behavior rule, and checking whether the task event satisfies the condition of the rule; S23. If the task event meets the conditions of the rule, the agent is commanded to execute the action corresponding to the rule and the processing of this task event is terminated.

8. A multi-agent collaborative task planning system, characterized in that: The system comprises: The asynchronous communication interface module is configured to build an asynchronous communication framework, respond to task events through an event-driven model, use the message buffer queue to decouple task event production and consumption, and realize asynchronous communication between the central controller and multiple agents; A rule engine module performs conditional matching on the task event based on predefined behavior rules, and if the match is successful, instructs the agent to execute the action corresponding to the rule; otherwise, the task event is marked as an unmatched task, and the unmatched task is transmitted to the LLM language model; An LLM decision-making auxiliary module is configured to load the historical process data of the unmatched task into the historical message context of the LLM language model, generate a task allocation decision in combination with a guidance template, and optimize the task allocation decision through an abnormal re-decision mechanism to obtain a decision result; The intelligent agent execution module is configured to distribute the decision results to the corresponding intelligent agents for execution, and to update the context data of the task event and the predefined behavior rules in real time.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the multi-agent collaborative task planning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the multi-agent collaborative task planning method as described in any one of claims 1 to 7 are implemented.

Images