CN-122022438-A - Complex-process-oriented industrial multi-agent collaboration system and method

CN122022438ACN 122022438 ACN122022438 ACN 122022438ACN-122022438-A

Abstract

The invention relates to the technical field of intelligent agent control, in particular to an industrial multi-intelligent agent cooperation system and method for a complex flow, which are used for generating a plurality of intelligent agent dispatching strategies by combining a multi-intelligent agent knowledge base and standby states of intelligent agents, generating a process flow characteristic library according to multi-source working condition data of an industrial site, defining a process flow dynamic partition according to the process flow dispatching strategies, determining a current intelligent agent allocation strategy of each partition according to the plurality of intelligent agent dispatching strategies and the process flow dynamic partition, determining conflict relations among the intelligent agents according to the operation states of the partitions, identifying scheduling characteristic labels of the intelligent agents, comparing the process flow execution vacancy characteristics of the adjacent partitions of the partitions with the scheduling characteristic labels of the intelligent agents of the partitions, and changing a part of the intelligent agents to serve as shared cooperation intelligent agents. By balancing the special and conditional scheduling of the intelligent agent in the process flow, the operation energy of the intelligent agent cluster is fully utilized, and the requirements of all the process flows on the intelligent agent are efficiently met.

Inventors

Wen Fangping
DONG QI
ZHAN PENG
LIU JUNJIE

Assignees

工业云制造(四川)创新中心有限公司

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. A complex flow-oriented industrial multi-agent collaboration system, comprising: The knowledge base generation module is used for acquiring the historical working records of the multi-agent clusters so as to generate a multi-agent knowledge base; the dispatching strategy generation module is used for generating a plurality of agent dispatching strategies according to the multi-agent knowledge base and the standby state of the agents; The process characteristic generating module is used for generating a process characteristic library according to multi-source working condition data of the industrial field; the partition defining module is used for defining the dynamic partition of the process flow according to the process flow feature library; the distribution strategy generation module is used for determining the current agent distribution strategy of each partition according to the agent dispatching strategies and the process flow dynamic partitions; the scheduling label determining module is used for determining conflict relations among the agents according to the operation live conditions of the partitions so as to identify scheduling characteristic labels of the agents; and the change module is used for comparing the process flows of other adjacent partitions of the partition with the execution vacancy characteristics of the process flows of the other adjacent partitions of the partition and the scheduling characteristic labels of the agents of the partition, and changing a part of the agents to be used as shared cooperative agents.
2. The complex flow-oriented industrial multi-agent collaboration system of claim 1, wherein: The knowledge base generating module is used for obtaining the historical working record of the multi-agent cluster so as to generate a multi-agent knowledge base, and comprises the following steps: Extracting an intrinsic working attribute and an extrinsic working attribute of each agent from a historical working record of a multi-agent cluster, wherein the intrinsic working attribute comprises a working program set in the agent; performing semantic recognition on the internal working attribute and the external working attribute to generate respective internal working labels and external working labels of all the intelligent agents, thereby forming a multi-intelligent agent knowledge base; the dispatch strategy generation module is used for generating a plurality of agent dispatch strategies according to the multi-agent knowledge base and the standby state of the agents, and comprises the following steps: the intelligent agent dispatching method comprises the steps of generating real-time standby state representation vectors of each intelligent agent according to real-time charging quantity and real-time positions of each intelligent agent, and generating a plurality of intelligent agent dispatching strategies according to the multi-intelligent agent knowledge base and the real-time standby state representation vectors, wherein the intelligent agent dispatching strategies comprise industrial operation scene attributes matched with the intelligent agents.
3. The complex flow-oriented industrial multi-agent collaboration system of claim 1, wherein: The process characteristic generating module is used for generating a process characteristic library according to multi-source working condition data of an industrial field, and comprises the following steps: Performing machine vision identification and hardware equipment detection on an industrial field to obtain multi-source working condition data of the industrial field, wherein the multi-source working condition data comprise respective construction and motion data of an object and hardware equipment; Performing time domain change identification on the multi-source working condition data to obtain respective construction change characteristics and motion change characteristics of an object and hardware equipment during each process flow, thereby generating a process flow characteristic library of the industrial field; the partition demarcation module is used for demarcating the dynamic partition of the process flow according to the process flow feature library, and comprises the following steps: And calculating the structural change feature similarity and the motion change feature similarity among all the process flows in the process flow feature library, so as to divide the industrial field into a plurality of process flow dynamic partitions, wherein each process flow dynamic partition corresponds to a plurality of process flows meeting the preset similarity condition.
4. The complex flow-oriented industrial multi-agent collaboration system of claim 1, wherein: the allocation policy generation module is configured to determine a current agent allocation policy of each partition according to the plurality of agent dispatch policies and the process flow dynamic partition, and includes: Comparing the plurality of agent dispatching strategies with all process flow dynamic partitions to obtain the completion efficiency of each agent dispatching strategy to all process flows subordinate to each process flow dynamic partition; determining a current agent allocation strategy of each partition according to the completion efficiency, wherein the agent allocation strategy comprises a corresponding relation between each partition and an agent currently allocated for executing a subordinate process flow of the agent; The scheduling tag determining module is configured to determine a conflict relationship between agents according to the operation condition of the partition, so as to identify a scheduling characteristic tag of the agent, and includes: The method comprises the steps of visually identifying operation live conditions of all distributed agents in each partition, obtaining execution action states of all distributed agents in corresponding process flows, determining conflict relations among all distributed agents, and determining respective schedulable time characteristics and schedulable work function characteristics of all distributed agents according to the conflict relations, so as to identify scheduling characteristic labels of the agents.
5. The complex flow-oriented industrial multi-agent collaboration system of claim 1, wherein: The change module is used for comparing the process flows of other adjacent partitions of the partition with the execution vacancy characteristics of the process flows of the other adjacent partitions of the partition and the scheduling characteristic labels of the agents of the partition, and changing a part of the agents to be shared collaborative agents comprises the following steps: visually identifying missing and/or error states of process flow execution of other adjacent partitions of the partition, thereby determining the missing feature of the process flow execution of the other partitions; And comparing the blank characteristic of the process flow execution with the scheduling characteristic label of the agents in the partition, so as to change a part of agents in the partition as shared cooperative agents, wherein the shared cooperative agents are configured to be capable of freely moving and switching between the partition and the other partitions to execute different process flows.
6. The complex flow-oriented industrial multi-agent cooperation method is characterized by comprising the following steps of: step S1, acquiring a historical work record of a multi-agent cluster to generate a multi-agent knowledge base; Step S2, generating a process flow feature library according to multi-source working condition data of an industrial field; Step S3, determining the current agent allocation strategy of each partition according to the agent dispatching strategies and the process flow dynamic partitions; according to the operation live condition of the subareas, determining conflict relation among the intelligent agents so as to identify scheduling characteristic labels of the intelligent agents; And S4, comparing the process flows of other adjacent partitions of the partition with the blank characteristic and the scheduling characteristic label of the agent of the partition, and changing a part of agents to be shared cooperative agents.
7. The complex flow-oriented industrial multi-agent collaboration method of claim 6, wherein: in step S1, a multi-agent knowledge base is generated by acquiring a historical work record of a multi-agent cluster, and a plurality of agent dispatching strategies are generated according to the multi-agent knowledge base and standby states of agents, including: Extracting an intrinsic working attribute and an extrinsic working attribute of each agent from a historical working record of a multi-agent cluster, wherein the intrinsic working attribute comprises a working program set in the agent; performing semantic recognition on the internal working attribute and the external working attribute to generate respective internal working labels and external working labels of all the intelligent agents, thereby forming a multi-intelligent agent knowledge base; the intelligent agent dispatching method comprises the steps of generating real-time standby state representation vectors of each intelligent agent according to real-time charging quantity and real-time positions of each intelligent agent, and generating a plurality of intelligent agent dispatching strategies according to the multi-intelligent agent knowledge base and the real-time standby state representation vectors, wherein the intelligent agent dispatching strategies comprise industrial operation scene attributes matched with the intelligent agents.
8. The complex flow-oriented industrial multi-agent collaboration method of claim 6, wherein: In step S2, a process flow feature library is generated according to multi-source working condition data of an industrial field, and a process flow dynamic partition is defined according to the process flow feature library, wherein the process flow dynamic partition comprises: Performing machine vision identification and hardware equipment detection on an industrial field to obtain multi-source working condition data of the industrial field, wherein the multi-source working condition data comprise respective construction and motion data of an object and hardware equipment; Performing time domain change identification on the multi-source working condition data to obtain respective construction change characteristics and motion change characteristics of an object and hardware equipment during each process flow, thereby generating a process flow characteristic library of the industrial field; And calculating the structural change feature similarity and the motion change feature similarity among all the process flows in the process flow feature library, so as to divide the industrial field into a plurality of process flow dynamic partitions, wherein each process flow dynamic partition corresponds to a plurality of process flows meeting the preset similarity condition.
9. The complex flow-oriented industrial multi-agent collaboration method of claim 6, wherein: In step S3, determining a current agent allocation policy of each partition according to the agent dispatch policies and the process flow dynamic partition, and determining a conflict relationship between agents according to the operation live of the partition, thereby identifying a scheduling characteristic label of the agent, including: Comparing the plurality of agent dispatching strategies with all process flow dynamic partitions to obtain the completion efficiency of each agent dispatching strategy to all process flows subordinate to each process flow dynamic partition; determining a current agent allocation strategy of each partition according to the completion efficiency, wherein the agent allocation strategy comprises a corresponding relation between each partition and an agent currently allocated for executing a subordinate process flow of the agent; The method comprises the steps of visually identifying operation live conditions of all distributed agents in each partition, obtaining execution action states of all distributed agents in corresponding process flows, determining conflict relations among all distributed agents, and determining respective schedulable time characteristics and schedulable work function characteristics of all distributed agents according to the conflict relations, so as to identify scheduling characteristic labels of the agents.
10. The complex flow-oriented industrial multi-agent collaboration method of claim 6, wherein: In step S4, comparing the process flows of the other adjacent partitions of the partition with the execution of the vacancy feature and the scheduling feature label of the agent of the partition, and changing a part of the agents as the shared cooperative agent, including: visually identifying missing and/or error states of process flow execution of other adjacent partitions of the partition, thereby determining the missing feature of the process flow execution of the other partitions; And comparing the blank characteristic of the process flow execution with the scheduling characteristic label of the agents in the partition, so as to change a part of agents in the partition as shared cooperative agents, wherein the shared cooperative agents are configured to be capable of freely moving and switching between the partition and the other partitions to execute different process flows.

Description

Complex-process-oriented industrial multi-agent collaboration system and method Technical Field The invention relates to the technical field of agent control, in particular to an industrial multi-agent cooperation system and method for complex flow. Background Robot and other intelligent agents are widely used in industrial scenes. The intelligent agent can be arranged in different working procedures in the industrial field, so that products on the corresponding working procedures are operated and treated, and the production efficiency and the yield of the products are improved. The application of the existing intelligent agents in the industrial scene is limited in a special mode, namely each intelligent agent is only responsible for the operation of a single process flow and cannot span to other process flows, the intelligent agents in different process flows cannot realize the coordination work of the process, so that the intelligent agent transportation energy of one part of the process flows is idle, the intelligent agent transportation energy of the other part of the process flows is tension, and the intelligent agent transportation energy cannot be fully utilized. Considering that the intelligent agent has stronger work executable and can realize more types of work tasks, if the intelligent agent is limited in a single process flow, the cooperation flexibility and the scheduling freedom of a plurality of intelligent agents in an industrial scene are reduced, and the requirements of all process flows on the intelligent agent cannot be met efficiently. Disclosure of Invention Considering that the existing multi-agent application in industrial scenes adopts a special mode, the agent can only be limited in a specific process flow, but cannot be freely scheduled in different process flows, so that the problems of unbalanced agent transportation energy distribution and incapability of flexible cooperation of the agent occur in different process flows. In view of the above problems, the present invention provides a complex flow-oriented industrial multi-agent collaboration system, including: The knowledge base generation module is used for acquiring the historical working records of the multi-agent clusters so as to generate a multi-agent knowledge base; the dispatching strategy generation module is used for generating a plurality of agent dispatching strategies according to the multi-agent knowledge base and the standby state of the agents; The process characteristic generating module is used for generating a process characteristic library according to multi-source working condition data of the industrial field; the partition defining module is used for defining the dynamic partition of the process flow according to the process flow feature library; the distribution strategy generation module is used for determining the current agent distribution strategy of each partition according to the agent dispatching strategies and the process flow dynamic partitions; the scheduling label determining module is used for determining conflict relations among the agents according to the operation live conditions of the partitions so as to identify scheduling characteristic labels of the agents; and the change module is used for comparing the process flows of other adjacent partitions of the partition with the execution vacancy characteristics of the process flows of the other adjacent partitions of the partition and the scheduling characteristic labels of the agents of the partition, and changing a part of the agents to be used as shared cooperative agents. Optionally, the knowledge base generating module is configured to obtain a historical working record of the multi-agent cluster, so as to generate a multi-agent knowledge base, and includes: Extracting an intrinsic working attribute and an extrinsic working attribute of each agent from a historical working record of a multi-agent cluster, wherein the intrinsic working attribute comprises a working program set in the agent; performing semantic recognition on the internal working attribute and the external working attribute to generate respective internal working labels and external working labels of all the intelligent agents, thereby forming a multi-intelligent agent knowledge base; the dispatch strategy generation module is used for generating a plurality of agent dispatch strategies according to the multi-agent knowledge base and the standby state of the agents, and comprises the following steps: the intelligent agent dispatching method comprises the steps of generating real-time standby state representation vectors of each intelligent agent according to real-time charging quantity and real-time positions of each intelligent agent, and generating a plurality of intelligent agent dispatching strategies according to the multi-intelligent agent knowledge base and the real-time standby state representation vectors, wherein the intelligent agent dispatching strategies c