Multi-agent collaborative decision-making method, system, equipment, medium and product

Abstract

The invention provides a multi-agent collaborative decision-making method, system and device, a medium and a product, and belongs to the field of business support, and the method comprises the steps: carrying out the registration and role distribution of multiple agents based on a member registration and election algorithm, and determining a round table member and a round table chairman; acquiring an external decision demand based on the round table chairman, selecting a round table member corresponding to the external decision demand as an expert group member, sending the external decision demand to the expert group member, and generating a proposal corresponding to the external decision demand by the expert group member; summarizing proposals fed back by the members of the expert group based on the round table chairman, distributing the summarized proposals to the members of the expert group, and voting the summarized proposals by the members of the expert group to obtain a voting result; voting results fed back by the members of each expert group are obtained based on the round table chairman, an optimal proposal is determined according to the voting results, and the optimal proposal is used as a decision scheme. The democratic votes are processed in parallel through multiple agents, the advantages of multi-agent decision making are exerted, and the decision making risk is reduced.

Description

Technical Field

[0001] This application relates to the field of business support, and in particular to a multi-agent collaborative decision-making method, system, equipment, medium and product. Background Art

[0002] With the rapid development of large-scale models and intelligent agent technologies, artificial intelligence is increasingly permeating core IT domains, such as operations and development. However, due to the varying quality of large models and the widespread problem of hallucinations, individual agents can exhibit bias in their decision-making, resulting in ineffective reliability of their results. In this context, improving the effectiveness of intelligent agents in IT and ensuring the stability and reliability of their decisions have become pressing challenges for the industry.

[0003] There are currently two main solutions for optimizing the reliability of agent decisions: Solution 1, multi-agent serial processing, connects agents from different domains and types to the same processing flow. Each agent handles a portion of the problem, and the last agent summarizes the output. This stacking of agents from different domains yields more reliable results. However, this solution only has one agent per domain, and the reliability issues of a single agent remain.

[0004] Solution 2, multi-agent star processing, uses a central agent node to interact with the outside world and distribute input to edge agent nodes for parallel processing. The central node aggregates all processing results and makes decisions. This approach achieves more reliable results through repeated calculations by multiple agents in the same domain. The decision reliability of this solution is limited by the central node. Although solution 1 addresses the reliability issues of single-agent decision making, the reliability of the decision depends largely on the model quality of the central agent node, which fails to leverage the advantages of multi-agent decision making. Summary of the Invention

[0005] The present application provides a multi-agent collaborative decision-making method, system, equipment, medium and product to solve the problem that the reliability of existing agent decision-making depends largely on the model quality of the central agent node and cannot give full play to the advantages of multi-agent decision-making.

[0006] This application provides a multi-agent collaborative decision-making method, including: Based on the member registration and election algorithm, multiple agents are registered and assigned roles to determine the roundtable members and roundtable chairperson; Based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; The roundtable chairperson summarizes the proposals fed back by the expert group members and distributes the summarized proposals to the expert group members. The expert group members vote on the summarized proposals to obtain voting results. The roundtable chair obtains the voting results of the expert group members based on the voting results, determines the best proposal based on the voting results, and uses the best proposal as the decision-making plan.

[0007] As an embodiment, obtaining the voting results of the expert group members based on the roundtable chair and determining the optimal proposal according to the voting results includes: Based on the voting results fed back by the expert group members obtained by the roundtable chair, a voting result matrix is ​​constructed according to the voting results; Obtaining a member weight configuration table based on the roundtable chairperson, and determining the member weight of each expert group member according to the member weight configuration table, wherein the member weight configuration table includes a static weight, a dynamic weighting factor, and a number of dynamic weight samples of each expert group member; Determine the voting score of each proposal based on the voting result matrix and the member weights; Based on the voting scores, it is determined whether there is a tie. If so, the aggregated proposals are redistributed to each of the expert group members for re-voting. If not, the proposal with the highest voting score is selected as the optimal proposal.

[0008] As an embodiment, determining the member weight of each expert group member according to the member weight configuration table includes: Determining the dynamic weight of each expert group member according to the ratio of the number of historical decision proposals of each expert group member to the number of dynamic weight samples; The member weight of each expert group member is determined according to the static weight, dynamic weighting factor and dynamic weight of each expert group member.

[0009] As an embodiment, the roundtable chair obtains external decision requirements, selects the roundtable members corresponding to the external decision requirements as expert group members, and sends the external decision requirements to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision requirements, including: Obtaining external decision requirements based on the roundtable chairperson, inputting the external decision requirements into the determination model, obtaining the business domain of the external decision requirements output by the determination model, and encapsulating the external decision requirements into an agenda message; The roundtable members corresponding to the business fields of the external decision-making needs are selected as expert group members, and the topic message is distributed to the expert group members. The expert group members are used to input the topic message into the proposal generation model to obtain the proposal corresponding to the external decision-making needs output by the proposal generation model.

[0010] The present application also provides a multi-agent collaborative decision-making system, including a multi-agent cluster; The multi-agent cluster is used to register and assign roles to multiple agents based on a member registration and election algorithm, and determine roundtable members and a roundtable chairperson; based on the roundtable chairperson obtaining external decision-making requirements, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; based on the roundtable chairperson summarizing the proposals fed back by each expert group member, and distributing the summarized proposals to each expert group member, the expert group members are used to vote on the summarized proposals to obtain voting results; based on the roundtable chairperson obtaining the voting results fed back by each expert group member, the optimal proposal is determined based on the voting results, and the optimal proposal is used as the decision-making solution.

[0011] As an embodiment, it also includes: The agent management and control module is used to maintain the cluster state of the multi-agent cluster, and the cluster state includes member registration and role assignment information and configuration information. The configuration information includes member weight configuration information, member field configuration information, election and proposal timeout time and voting processing plan.

[0012] As an embodiment, it also includes: The operation analysis module is used to obtain, process and store the performance data and business data of each intelligent agent. The performance data includes the proposal delay, current load, proposal success rate and throughput of each intelligent agent. The business data includes the proposal record, proposal voting record, proposal selection record of each intelligent agent, and also includes the external decision-making requirements received by the multi-agent cluster and the decision-making plan output.

[0013] The present application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the multi-agent collaborative decision-making method as described above is implemented.

[0014] The present application also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements any of the multi-agent collaborative decision-making methods described above.

[0015] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the multi-agent collaborative decision-making methods described above.

[0016] The multi-agent collaborative decision-making method, system, equipment, medium and product provided in this application are based on the interaction between the roundtable chairman and the user, obtaining the external decision-making requirements input by the user, distributing the external decision-making requirements to the expert group members so that the expert group members can process them in parallel, generating proposals, and summarizing the proposals based on the roundtable chairman and distributing them again to the expert group members so that the expert group members can vote. The roundtable chairman determines the optimal proposal based on the voting results, and through multi-agent parallel processing of democratic voting, the advantages of multi-agent decision-making are brought into play and the decision-making risks are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 It is a flow chart of the multi-agent collaborative decision-making method provided in this application.

[0019] Figure 2 This is a flowchart of the member registration and election process provided by this application.

[0020] Figure 3 This is a flowchart of the topic setting and allocation provided by this application.

[0021] Figure 4 This is a flowchart of the proposal generation and submission process provided by this application.

[0022] Figure 5 This is a flowchart of democratic voting and decision-making provided by this application.

[0023] Figure 6 It is a structural diagram of the multi-agent collaborative decision-making system provided by this application.

[0024] Figure 7 This is a business process diagram of the registration and election module provided by this application.

[0025] Figure 8 This is a business process diagram for member registration provided by this application.

[0026] Figure 9 This is a business process diagram for the chairman election provided by this application.

[0027] Figure 10 This is a business process diagram of the topic proposal module provided by this application.

[0028] Figure 11 This is a business process diagram of the agenda setting provided by this application.

[0029] Figure 12 This is a business process diagram of topic allocation provided by this application.

[0030] Figure 13 This is a business process diagram of the topic proposal provided in this application.

[0031] Figure 14 This is a business process diagram of the voting decision module provided by this application.

[0032] Figure 15 This is a business process diagram of the intelligent body management and control module provided in this application.

[0033] Figure 16 This is a business process diagram of the operation analysis module provided by this application.

[0034] Figure 17 It is a structural diagram of the electronic device provided in this application. DETAILED DESCRIPTION

[0035] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0036] It should be noted that all actions of acquiring signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the location and with the authorization given by the owner of the corresponding device.

[0037] While multi-agent serial processing can connect multiple agents across different domains to meet decision-making needs in a macro sense, at the micro level, only one agent in each domain can handle the task, thus failing to address the reliability issues inherent in a single agent. Furthermore, due to the serial nature of this model, overall processing performance is significantly limited.

[0038] In the multi-agent star processing solution, although parallel processing provides more decision options, the final decision is made by the central node. This means that the reliability of the system decision depends largely on the model quality of the central agent node, and the advantages of multi-agent system decision-making cannot be fully utilized.

[0039] It can be seen that the existing multi-agent decision-making scheme has not achieved the inherent requirement of ensuring the reliability of agent decision-making through multi-agent collaboration, and cannot meet the needs of high-quality and high-reliability agent decision-making in the IT field. To this end, this application provides a multi-agent collaborative decision-making method, system, equipment, medium and product to improve the overall decision-making quality and reliability of the system, which is explained in detail below in conjunction with the drawings in the specification.

[0040] Figure 1 This is a flowchart of the multi-agent collaborative decision-making method provided by this application, such as Figure 1 As shown, the present application provides a multi-agent collaborative decision-making method. Multi-agents generally refer to multiple agents that exist and interact in parallel in the same environment. Agents can be robots, software programs, or any entity that can perceive the environment and take action. In the embodiment of the present application, multi-agents include agents in the IT field, and each agent can be distributed or centrally deployed. Specifically, the multi-agent collaborative decision-making method includes steps S100 to S400.

[0041] Step S100: Based on the member registration and election algorithm, multiple agents are registered and assigned roles to determine the round table members and the round table chairperson.

[0042] In step S200, based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members. The external decision-making requirements are then sent to the expert group members, who are then instructed to generate proposals corresponding to the external decision-making requirements. After generating proposals corresponding to the external decision-making requirements, the expert group members submit the proposals to the roundtable chair.

[0043] In step S300, the roundtable chairperson summarizes the proposals fed back by the expert group members and distributes the summarized proposals to the expert group members. The expert group members vote on the summarized proposals to obtain voting results.

[0044] Step S400: The roundtable chair obtains the voting results of the expert group members based on their feedback, and determines the best proposal based on the voting results, and uses the best proposal as the decision-making solution.

[0045] This application proposes a roundtable algorithm for multi-agent collaborative decision-making. This algorithm improves the overall decision-making quality and reliability of multi-agents through the principles of "parallel processing and democratic voting." The roundtable algorithm consists of four steps: member registration and election, agenda setting and allocation, proposal generation and submission, and democratic voting and decision-making. Member registration and election includes step S100, agenda setting and allocation, and proposal generation and submission include step S200, and democratic voting and decision-making includes steps S300-S400.

[0046] like Figure 2 As shown, the roundtable algorithm proposed in this application divides multiple agents into domain agents and general agents according to their functions. Domain agents refer to agents that include large models of specific fields (such as the IT field), and general agents refer to agents that include large models of general fields. New agents need to register and are divided into roundtable members and roundtable chairmen according to their functions. Optionally, domain agents are roundtable members, and general agents can participate in the chair election to become roundtable chairmen. When the roundtable chair loses contact, the general agent initiates the chair election and becomes a candidate. After receiving the election request, all agents (including candidates) vote and select candidates with larger weights. When a candidate receives more than half of the votes, he or she is elected as the chair and broadcasts it through heartbeats. After receiving the heartbeats, all agents update the roundtable status.

[0047] It is understandable that based on the interaction between the roundtable chair and the user, the external decision-making requirements input by the user are obtained, the external decision-making requirements are distributed to the expert group members so that the expert group members can process them in parallel and generate proposals. Based on the roundtable chair, the proposals are summarized and distributed to the expert group members again so that the expert group members can vote. The roundtable chair determines the optimal proposal based on the voting results. Through multi-agent parallel processing and democratic voting, the advantages of multi-agent decision-making are brought into play and the decision-making risks are reduced. This application proposes and adopts a roundtable algorithm based on a distributed architecture. Through parallel processing and democratic voting, it solves the problems of unreliable decision-making and high dependence on central nodes caused by single-agent decision-making in the traditional model. While ensuring the overall decision-making performance of the multi-agent system, it greatly improves the quality and reliability of decision-making.

[0048] Based on the above embodiment, as an optional embodiment, the roundtable chair obtains external decision-making requirements, selects the roundtable members corresponding to the external decision-making requirements as expert group members, and sends the external decision-making requirements to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements, including steps S210-S220.

[0049] Step S210 , based on the roundtable chairperson, obtains external decision requirements, inputs the external decision requirements into the determination model, obtains the business field of the external decision requirements output by the determination model, and encapsulates the external decision requirements into an agenda message.

[0050] In step S220, the roundtable members corresponding to the business areas of the external decision-making requirements are selected as expert group members. The topic message is distributed to the expert group members. The expert group members input the topic message into a proposal generation model, and the proposal generation model outputs a proposal corresponding to the external decision-making requirements. The proposal generation model can be constructed based on a large language model (LLM).

[0051] like Figure 3 As shown in the figure, the roundtable chair is responsible for receiving external decision-making needs initiated by users and setting them as topics. After receiving the topic, he first determines the field to which the topic belongs, then finds members in the corresponding field to form an expert group, and distributes the topic to the expert group members. After receiving it, the expert group members reply to confirm.

[0052] like Figure 4 As shown in the figure, after receiving the topic, each expert group member calls their own large model to generate a proposal and submits it to the roundtable chairman. The roundtable chairman confirms that the proposal has been received after confirming that the submitter and the topic of the proposal message are correct.

[0053] like Figure 5 As shown, the roundtable chair will summarize the proposals received and redistribute them to the expert panel members. Each expert panel member will then vote to select a preferred proposal. The roundtable chair will then confirm the votes. Similarly, a timeout mechanism is implemented during the proposal voting process, requiring expert panel members to complete their votes within a specified timeframe. After receiving the votes, the roundtable chair will tally the votes based on member weights and historical performance, and the proposal with the highest number of votes will be adopted. In the event of a tie, a re-vote will be conducted based on the configuration, or the roundtable chair will make the decision. Finally, the roundtable chair will communicate the decision to the external community.

[0054] It can be understood that this application uses a judgment model to determine the business field of external decision-making needs, encapsulates the external decision-making needs into topic messages, and distributes them to expert group members corresponding to the business field, so that expert group members can generate proposals according to their respective proposal generation models, realize parallel processing, and improve proposal generation efficiency and accuracy.

[0055] Based on the above embodiment, as an optional embodiment, the voting results of the expert group members are obtained based on the roundtable chair, and the optimal proposal is determined according to the voting results, including steps S410 to S440.

[0056] Step S410: The roundtable chair obtains voting results fed back by each expert group member, and constructs a voting result matrix according to the voting results.

[0057] In the voting result matrix, the voting results of each expert group member can be described by a matrix. The rows in the matrix identify proposals p, and the columns identify members m. In the matrix cells, 1 indicates a vote in favor and 0 indicates a vote against, as shown in the following table.

[0058] Step S420: Based on the member weight configuration table obtained by the roundtable chairman, the member weight of each expert group member is determined according to the member weight configuration table. The member weight configuration table includes the static weight, dynamic weighting factor and dynamic weight sample number of each expert group member, which can adapt to the characteristics of different expert group members. The static weight is a feature that identifies the general ability of the member based on experience, the dynamic weighting factor is used to identify the dynamic weight ratio of the member, and the dynamic weight sample number is used to identify the number of samples required for dynamic weight statistics of the member.

[0059] The member weight configuration table is shown in the following table.

[0060] Optionally, determining the member weight of each expert group member according to the member weight configuration table includes steps S421 and S422.

[0061] Step S421 : determining the dynamic weight of each expert group member according to the ratio of the number of historical decision proposals of each expert group member to the number of dynamic weight samples.

[0062] Step S422: Determine the member weight of each expert group member according to the static weight, dynamic weighting factor and dynamic weight of each expert group member.

[0063] The weight of a member needs to consider both static weight and dynamic weight. The static weight comes from the configuration settings, while the dynamic weight is measured based on the member's historical performance, that is, the proportion of proposals selected by the decision-maker in the number of dynamic weight samples. The final member weight is calculated by combining the dynamic weighting factors.

[0064] in, For the i The dynamic weight of each expert group member, For the i The number of proposals selected by the expert group members, For thei The number of dynamic weight samples of expert group members, For the i Dynamic weighting factors of each expert group member, For the i The member weight of each expert group member.

[0065] Step S430: Determine the voting score of each proposal based on the voting result matrix and the member weights.

[0066] Based on the member weights and voting matrix, the voting score of each proposal in the voting result matrix can be calculated. The final voting score of each column, i.e. each proposal, can be calculated by summing up. The one with the largest score is regarded as the optimal solution to form a decision.

[0067] in, The voting result matrix j The proposal of the column, The i-th expert group member j Column proposal voting results, vote for or against (0 or 1), n is the number of expert group members.

[0068] Step S440: Determine whether there is a tie based on the voting scores. If so, redistribute the aggregated proposals to each expert group member for re-voting. If not, take the proposal with the highest voting score as the optimal proposal.

[0069] It can be understood that this application uses the historical performance of the intelligent agent as part of the weight measurement and calculates votes and scores based on a matrix weighted algorithm, thereby amplifying the weight of high-quality intelligent agents in the overall decision-making, and realizing dynamic adjustment and positive feedback loop of the overall decision-making quality of the intelligent agent cluster.

[0070] The multi-agent collaborative decision-making system provided by the present application is described below. The multi-agent collaborative decision-making system described below and the multi-agent collaborative decision-making method described above can be referenced to each other.

[0071] Figure 6 This is a schematic diagram of the structure of the multi-agent collaborative decision-making system provided by this application. Figure 6 As shown, the present application also provides a multi-agent collaborative decision-making system, including a multi-agent cluster 610.

[0072] The multi-agent cluster 610 is used to register and assign roles to multiple agents based on a member registration and election algorithm, and determine roundtable members and a roundtable chairperson; based on the roundtable chairperson obtaining external decision-making requirements, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; based on the roundtable chairperson summarizing the proposals fed back by each expert group member, and distributing the summarized proposals to each expert group member, the expert group members are used to vote on the summarized proposals to obtain voting results; based on the roundtable chairperson obtaining the voting results fed back by each expert group member, the optimal proposal is determined based on the voting results, and the optimal proposal is used as the decision-making solution.

[0073] As an embodiment, the multi-agent collaborative decision-making system provided by this application further includes: The agent management and control module 620 is used to maintain the cluster state of the multi-agent cluster, and the cluster state includes member registration and role assignment information and configuration information. The configuration information includes member weight configuration information, member field configuration information, election and proposal timeout time and voting processing plan.

[0074] As an embodiment, the multi-agent collaborative decision-making system provided by this application further includes: The operation analysis module 630 is used to obtain, process and store the performance data and business data of each intelligent agent. The performance data includes the proposal delay, current load, proposal success rate and throughput of each intelligent agent. The business data includes the proposal record, proposal voting record, proposal selection record of each intelligent agent, and also includes the external decision-making requirements received by the multi-agent cluster and the decision-making plan output.

[0075] The multi-agent collaborative decision-making system provided in this application includes a multi-agent cluster, an agent management and control module, and an operation analysis module. Each agent includes a registration and election module, an issue proposal module, and a voting decision module, which are mainly responsible for the construction of the multi-agent cluster and the processing of decision-making needs. The agent management and control module is responsible for the status maintenance and configuration management of the cluster, and provides capabilities such as member information registration and update, cluster status change and notification. The operation analysis module is mainly responsible for recording, storing, and analyzing the operation of the cluster, including the performance of the cluster and members, the records of proposals and voting, the historical performance of each member, etc., and continuously optimizing the decision-making quality of the cluster.

[0076] like Figure 7 As shown in the figure, the registration and election module is mainly responsible for the construction of the multi-agent cluster, including member registration, chairman election, and election voting.

[0077] Member registration: Member registration is the first step in building a multi-agent cluster. After the agent is started, it automatically initiates a registration request to the agent control module. The agent control module registers and initializes the configuration information of the agent and confirms the registration success. The agent control module then updates the cluster status and sends cluster change messages to all members of the cluster. Members update the cluster status cache. After receiving the confirmation of successful registration, the new member starts to listen to the messages of the control module and other members. The specific business process is as follows Figure 8 shown.

[0078] Chairman election: A multi-agent cluster requires a roundtable chairman to implement demand monitoring and decision-making tasks. Therefore, when the current roundtable chairman loses connection due to device or network problems (heartbeat timeout), all general agents in the cluster will initiate a chairman election and send election messages to other agents. For detailed business processes, see Figure 9 .

[0079] Topic Proposal Module: The roundtable chair is mainly responsible for monitoring external decision-making needs and handing them over to the intelligent agents in the corresponding fields to generate solutions, including three parts: topic setting, topic allocation, and topic proposal. For specific business processes, see Figure 10 .

[0080] Agenda setting: In a multi-agent cluster, the roundtable chair is responsible for communicating with external users and monitoring their decision-making needs. After receiving the needs, the roundtable chair will use its own model capabilities to determine the business area to which the needs belong and encapsulate them into an agenda message. For the specific business process, see Figure 11 .

[0081] Topic assignment: After the topic is set, the roundtable chair will find members with the same field and low load in the cluster to form an expert group based on the topic area, and then assign the topic to the expert group members through messages, so that they can generate proposals. For the specific business process, see Figure 12 .

[0082] Topic proposal: After receiving the topic, each expert group member uses their own big model to generate a proposal and submits it to the roundtable chairman. The roundtable chairman confirms that the proposal has been received after confirming that the submitter and topic are correct. For the specific business process, see Figure 13 .

[0083] Voting decision module: It is mainly responsible for reviewing and voting on proposals, selecting the best proposal and returning it to users. It mainly includes three parts: proposal distribution, proposal voting, and proposal decision. For the specific business process, see Figure 14 .

[0084] Proposal distribution: The roundtable chair will organize all proposals and distribute them to the expert group members, who will review and vote on each proposal.

[0085] Proposal voting: After receiving the proposals, the expert group members use the big model to evaluate the proposals separately, vote for the proposal they think is the best, and submit the voting results to the roundtable chair.

[0086] Chair Decision: After receiving proposal votes, the roundtable chair counts the votes and, based on the results, selects the optimal proposal to be returned to the external community. In the event of a tie, a re-vote or a proposal by the roundtable chair is determined based on the configuration. To account for the varying capabilities and characteristics of different agents, the voting rules incorporate the weights and historical performance of each member, using a weighted algorithm.

[0087] The above three modules with different functions are the core implementation of the "Round Table Algorithm" and are also key components of the cluster construction and decision-making process in the multi-agent collaborative decision-making system. The system provided in this application has the typical characteristics of a distributed system and has good scalability, fault tolerance, flexibility and reliability. In order to achieve system management and operation, the system can also be configured, monitored, analyzed and optimized with the agent management module and the operation analysis module to improve the decision-making performance and quality of the multi-agent collaborative decision-making system.

[0088] The agent management and control module is responsible for maintaining the cluster status of the multi-agent collaborative decision-making system and provides capabilities such as member registration, configuration management, and status notification. Among them, member registration is responsible for registering new member information and notifying existing members of the cluster. Configuration management is responsible for configuring the key processes of cluster business processing and regulating the decision-making performance and quality of the cluster. Common configurations include member weights, member fields, election and proposal timeouts, and tie vote processing methods. Status notification is responsible for notifying all members to update the cache through messages after the cluster status changes due to new member registration or configuration updates. For specific business processes, see Figure 15 .

[0089] The operation analysis module is responsible for collecting, processing, and storing the performance data and business data of the intelligent agents. The performance data includes the proposal latency, current load, success rate, throughput, etc. of the intelligent agents. The business data includes the proposal records, proposal voting records, proposal selection records, and the demand input and decision output records received by the cluster as a whole. The data is stored in the data warehouse by category. Statistical reports are used to assist operators in analyzing the operation of the cluster. Historical performance queries are used to assist the cluster chairman in calculating the dynamic weight of an intelligent agent, thereby improving the overall decision-making efficiency and quality of the cluster. For specific business processes, please refer to Figure 16 .

[0090] In summary, this application adopts a round-table algorithm with a distributed architecture. Through parallel processing and democratic voting, while taking into account the cluster decision-making performance, it greatly improves the decision-making quality and reliability of the multi-agent collaborative decision-making system, and completely solves the low performance and low quality problems brought about by traditional serial processing. The collective decision-making of multi-agent voting replaces the single decision under the traditional star structure, improving the reliability and credibility of the decision. At the same time, the distributed design has higher scalability and reliability than the centralized design under the star structure.

[0091] Figure 17 An example of a physical structure diagram of an electronic device is shown below. Figure 17 As shown, the electronic device may include: a processor 1710, a communications interface 1720, a memory 1730, and a communication bus 1740, wherein the processor 1710, the communications interface 1720, and the memory 1730 communicate with each other via the communication bus 1740. The processor 1710 may call the logic instructions in the memory 1730 to execute the multi-agent collaborative decision-making method, which includes: Based on the member registration and election algorithm, multiple agents are registered and assigned roles to determine the roundtable members and roundtable chairperson; Based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; The roundtable chairperson summarizes the proposals fed back by the expert group members and distributes the summarized proposals to the expert group members. The expert group members vote on the summarized proposals to obtain voting results. The roundtable chair obtains the voting results of the expert group members based on the voting results, determines the best proposal based on the voting results, and uses the best proposal as the decision-making plan.

[0092] In addition, the logical instructions in the above-mentioned memory 1730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program code.

[0093] On the other hand, the present application further provides a computer program product, comprising a computer program, which may be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the multi-agent collaborative decision-making method provided by the above methods, which includes: Based on the member registration and election algorithm, multiple agents are registered and assigned roles to determine the roundtable members and roundtable chairperson; Based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; The roundtable chairperson summarizes the proposals fed back by the expert group members and distributes the summarized proposals to the expert group members. The expert group members vote on the summarized proposals to obtain voting results. The roundtable chair obtains the voting results of the expert group members based on the voting results, determines the best proposal based on the voting results, and uses the best proposal as the decision-making plan.

[0094] In another aspect, the present application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the multi-agent collaborative decision-making method provided by the above methods, the method comprising: Based on the member registration and election algorithm, multiple agents are registered and assigned roles to determine the roundtable members and roundtable chairperson; Based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; The roundtable chairperson summarizes the proposals fed back by the expert group members and distributes the summarized proposals to the expert group members. The expert group members vote on the summarized proposals to obtain voting results. The roundtable chair obtains the voting results of the expert group members based on the voting results, determines the best proposal based on the voting results, and uses the best proposal as the decision-making plan.

[0095] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0096] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-agent collaborative decision-making method, characterized in that: include: Based on the member registration and election algorithm, multiple agents are registered and assigned roles to determine the roundtable members and roundtable chairperson; Based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; The roundtable chairperson summarizes the proposals fed back by the expert group members and distributes the summarized proposals to the expert group members. The expert group members vote on the summarized proposals to obtain voting results. The roundtable chair obtains the voting results of the expert group members based on the voting results, determines the best proposal based on the voting results, and uses the best proposal as the decision-making plan.

2. The multi-agent collaborative decision-making method according to claim 1, characterized in that: The roundtable chair obtains the voting results of the expert group members' feedback and determines the optimal proposal based on the voting results, including: Based on the voting results fed back by the expert group members obtained by the roundtable chair, a voting result matrix is ​​constructed according to the voting results; Obtaining a member weight configuration table based on the roundtable chairperson, and determining the member weight of each expert group member according to the member weight configuration table, wherein the member weight configuration table includes a static weight, a dynamic weighting factor, and a number of dynamic weight samples of each expert group member; Determine the voting score of each proposal based on the voting result matrix and the member weights; Based on the voting scores, it is determined whether there is a tie. If so, the aggregated proposals are redistributed to each of the expert group members for re-voting. If not, the proposal with the highest voting score is selected as the optimal proposal.

3. The multi-agent collaborative decision-making method according to claim 2, characterized in that: Determining the member weight of each expert group member according to the member weight configuration table includes: Determining the dynamic weight of each expert group member according to the ratio of the number of historical decision proposals of each expert group member to the number of dynamic weight samples; The member weight of each expert group member is determined according to the static weight, dynamic weighting factor and dynamic weight of each expert group member.

4. The multi-agent collaborative decision-making method according to claim 1, characterized in that: The process of obtaining an external decision-making requirement based on the roundtable chairperson, selecting the roundtable members corresponding to the external decision-making requirement as expert group members, and sending the external decision-making requirement to the expert group members, so that the expert group members generate proposals corresponding to the external decision-making requirement, includes: Obtaining external decision requirements based on the roundtable chairperson, inputting the external decision requirements into the determination model, obtaining the business domain of the external decision requirements output by the determination model, and encapsulating the external decision requirements into an agenda message; The roundtable members corresponding to the business fields of the external decision-making needs are selected as expert group members, and the topic message is distributed to the expert group members. The expert group members are used to input the topic message into the proposal generation model to obtain the proposal corresponding to the external decision-making needs output by the proposal generation model.

5. A multi-agent collaborative decision-making system, characterized in that: including multi-agent swarms; The multi-agent cluster is used to register and assign roles to multiple agents based on a member registration and election algorithm, and determine round table members and round table chairmen; Based on the external decision-making requirements obtained by the roundtable chair, the roundtable members corresponding to the external decision-making requirements are selected as expert group members, and the external decision-making requirements are sent to the expert group members, and the expert group members are used to generate proposals corresponding to the external decision-making requirements; based on the proposals fed back by the expert group members, the roundtable chair summarizes the proposals and distributes the summarized proposals to the expert group members, and the expert group members are used to vote on the summarized proposals to obtain voting results; The roundtable chair obtains the voting results of the expert group members based on the voting results, determines the best proposal based on the voting results, and uses the best proposal as the decision-making plan.

6. The multi-agent collaborative decision-making system according to claim 5, characterized in that: Also includes: The agent management and control module is used to maintain the cluster state of the multi-agent cluster, and the cluster state includes member registration and role assignment information and configuration information. The configuration information includes member weight configuration information, member field configuration information, election and proposal timeout time and voting processing plan.

7. The multi-agent collaborative decision-making system according to claim 5, characterized in that: Also includes: The operation analysis module is used to obtain, process and store the performance data and business data of each intelligent agent. The performance data includes the proposal delay, current load, proposal success rate and throughput of each intelligent agent. The business data includes the proposal record, proposal voting record, proposal selection record of each intelligent agent, and also includes the external decision-making requirements received by the multi-agent cluster and the decision-making plan output.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the multi-agent collaborative decision-making method as described in any one of claims 1 to 4 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the multi-agent collaborative decision-making method as described in any one of claims 1 to 4.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the multi-agent collaborative decision-making method as described in any one of claims 1 to 4.

Images