Task orchestration engine implementation method based on multiple agents

Abstract

The invention relates to the technical field of artificial intelligence and industrial automation, in particular to a task arrangement engine implementation method based on multiple agents. The method comprises the following steps: constructing a multi-agent cooperation framework which covers multiple types of agents such as a control agent, a diagnosis agent and a maintenance agent; a layered task decomposition mechanism is adopted to divide a complex business task into three levels of sub-task flows; the method comprises the following steps: constructing an ordered collaboration process among intelligent agents based on a state machine model, wherein the five states of definition readiness, diagnosis, scheme formulation and execution are completed; and multi-agent collaborative operation is realized through a collaborative network based on message passing. According to the task arrangement engine, efficient diagnosis and response of equipment faults can be achieved, complex service logic is completed through cooperation of the intelligent agents, and the technical problems that the processing capacity of a traditional single intelligent agent is limited, and multi-task coordination is difficult are solved.

Description

Technical Field

[0001] This invention belongs to the field of interdisciplinary technology of artificial intelligence and industrial automation, and specifically relates to a method for implementing a task orchestration engine based on multi-agent intelligence, which is particularly suitable for scenarios such as fault diagnosis and intelligent operation and maintenance of industrial equipment that require multi-agent collaborative processing of complex business processes. Background Technology

[0002] With the development of Industry 4.0 and intelligent manufacturing, equipment fault diagnosis and maintenance are becoming increasingly complex. Traditional fault diagnosis systems typically employ a single model or rule engine, which suffers from the following technical shortcomings: First, the processing capacity of a single agent is limited, making it difficult to handle complex tasks such as multi-source data fusion and multi-parameter correlation analysis; second, traditional systems lack effective task coordination mechanisms, making collaboration between multiple processing modules difficult, easily leading to information silos and decision conflicts; furthermore, existing systems lack flexible state management and anomaly recovery mechanisms, often failing to automatically adjust strategies when anomalies occur in the processing flow, resulting in task interruptions or error accumulation. Although multi-agent-based systems exist in the current technology, their integration and intelligence levels in task orchestration, state management, and anomaly handling still need improvement. Therefore, there is an urgent need in this field for a task orchestration engine that can effectively coordinate multiple agents, achieve orderly execution of complex tasks, and possess adaptive capabilities. Summary of the Invention

[0003] The main technical problem this invention aims to solve is overcoming the limitations of traditional fault diagnosis systems, such as the limited capabilities of a single agent, difficulties in multi-task coordination, and a lack of adaptive adjustment capabilities. To address these problems, this invention proposes a multi-agent-based task orchestration engine implementation method. This method achieves efficient collaborative processing of complex tasks through techniques such as constructing a multi-agent cooperation framework, hierarchical task decomposition, and state machine-driven cooperation. Another objective of this invention is to provide a task orchestration engine system that implements the above method. The beneficial technical effects of this invention include: 1. Enhanced ability to handle complex tasks: Through the division of labor and cooperation among multiple intelligent agents, it can handle complex business scenarios that require multi-source data analysis and multi-step decision-making.

[0004] 2. Enhanced system collaboration efficiency: The state machine-based ordered collaboration process ensures that each agent performs the right task at the right time, avoiding conflicts and duplication.

[0005] 3. Improved system continuity: When a task execution fails, the system can automatically roll back the state and switch the processing mode to ensure the continuity of business processes.

[0006] 4. Knowledge sharing and reuse are realized: Knowledge sharing among intelligent agents is realized through context storage, which improves the efficiency and quality of problem solving. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the architecture of a task orchestration engine system in one embodiment of the present invention.

[0008] Figure 2 This is a flowchart of a multi-agent-based task orchestration method in one embodiment of the present invention.

[0009] Figure 3 This is a schematic diagram of the collaborative process of the intelligent agent for equipment fault diagnosis in one embodiment of the present invention. Detailed Implementation The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0010] refer to Figure 1 This demonstrates the overall architecture of a multi-agent task orchestration engine system. The system adopts a modular design, including a multi-agent collaboration module, a hierarchical task decomposition module, a state machine-driven module, a message passing network module, and an exception handling module. These modules work together to form a complete task processing closed loop.

[0011] refer to Figure 2 The multi-agent-based task orchestration method of the present invention includes the following steps: Step S201 (Multi-Agent Framework Construction): Construct a collaborative framework comprising multiple types of agents, including control agents, diagnostic agents, and maintenance agents. The control agent is responsible for overall process control and coordination; the diagnostic agent is responsible for fault analysis and problem diagnosis; and the maintenance agent is responsible for developing repair plans and guiding maintenance operations. Each agent is built on an open-source technology stack and possesses specific professional capabilities.

[0012] Step S202 (Hierarchical Task Decomposition): A hierarchical task decomposition mechanism is adopted to divide complex business tasks into three levels of sub-task flows. The first-level task is to collect data such as vibration and temperature in real time by the equipment sensors through the OPC UA protocol. When the monitored value exceeds the preset threshold, a fault event is triggered. The second-level task is executed by the diagnostic agent, which includes two parallel sub-tasks: "historical work order retrieval" and "multi-parameter correlation analysis". Multi-threaded processing is used to improve efficiency. The third-level task is handled by the maintenance agent, which generates a detailed operation sequence based on the diagnostic conclusions, including maintenance steps, required spare parts, safety precautions, etc.

[0013] Step S203 (State Machine-Driven Collaboration): Based on the LangGraph state machine model, construct an ordered collaborative process between agents, defining five states: Ready, Diagnosing, Solution Development, Executing, and Completed. When equipment vibration anomalies, the system transitions from the "Ready" state to the "Diagnosing" state, activating the diagnostic agent to begin work. After diagnosis, the state transitions to "Solution Development," where the maintenance agent retrieves the maintenance manual and confirms spare parts availability. The control agent issues maintenance commands to the PLC and then enters the "Executing" state. Once sensor feedback data returns to normal, the task is marked as "Completed."

[0014] Step S204 (Message Passing Collaboration): Multi-agent collaborative operation is achieved through a message passing-based collaborative network. Device sensors use the MQTT protocol to publish fault events in real time, triggering the control agent to initiate the diagnostic process; the diagnostic agent and maintenance agent share knowledge and synchronize states through context storage. The message passing mechanism ensures that each agent can obtain the necessary information in a timely manner and coordinate its actions.

[0015] Step S205 (Exception Handling and Recovery): When an exception occurs during task execution, the system state machine automatically rolls back and switches the processing mode. For example, when the PLC command fails to be issued, the system state machine automatically rolls back to the "Solution Formulation" node, the maintenance agent switches to remote guidance mode, and pushes real-time maintenance guidance to on-site personnel using AR technology to ensure the continuity and controllability of the operation process under abnormal conditions.

[0016] Throughout the process, each intelligent agent coordinates its work rhythm through state machines and shares information and knowledge through message passing, forming an efficient collaborative processing mechanism. The system particularly emphasizes the "human-in-the-loop" design, allowing maintenance personnel to intervene in the process at any time, providing professional judgment or performing manual operations, thus forming a human-machine collaborative intelligent operation and maintenance system.

[0017] This patent application fully elaborates on the technological innovations of a multi-agent task orchestration engine, particularly the core innovations such as the multi-agent collaborative framework, hierarchical task decomposition, state machine-driven collaboration, and exception handling mechanism. These innovations have significant technological advancements and practical value in the fields of industrial equipment fault diagnosis and intelligent operation and maintenance.

Claims

1. A method for implementing a multi-agent-based task orchestration engine, characterized in that, Includes the following steps: S1: Construct a multi-agent collaborative framework, including a control agent, a diagnostic agent, and a maintenance agent; S2: Adopt a hierarchical task decomposition mechanism to divide complex business tasks into three levels of sub-task flows; S3: Construct an orderly collaborative process among agents based on a state machine model and define the task execution state; S4: Achieve multi-agent collaborative operation through a message-passing-based collaborative network; S5: When a task execution error occurs, the system state machine automatically rolls back and switches the processing mode. The method according to claim 1, characterized in that, The "hierarchical task decomposition mechanism" in step S2 specifically includes: Level 1 tasks are data collected in real time by the equipment sensors and fault events are triggered; Level 2 tasks are parallel subtasks of historical work order retrieval and multi-parameter correlation analysis performed by the diagnostic agent; Level 3 tasks are operation sequences generated by the maintenance agent based on the diagnostic conclusions. The method according to claim 1, characterized in that, The "state machine model" in step S3 includes: ready state: the system waits for a fault event to be triggered; diagnostic state: the diagnostic agent performs fault analysis; solution formulation state: the maintenance agent formulates a maintenance plan; execution state: the control agent issues execution instructions; completed state: the task is completed and the result is confirmed. A task orchestration engine system for implementing the method of any one of claims 1-3, characterized in that, include: The multi-agent collaboration module includes a control agent, a diagnostic agent, and a maintenance agent; The hierarchical task decomposition module is used to divide complex tasks into multi-level subtask flows; The state machine-driven module builds collaborative processes based on the LangGraph state machine model. The message passing network module enables communication between intelligent agents via the MQTT protocol; The exception handling module supports state rollback and mode switching functions.

Images