Multi-agent-based product research and development decision optimization method and system

Abstract

The invention discloses a multi-agent-based product research and development decision optimization method and system, and belongs to the technical field of artificial intelligence. The invention aims to solve the technical problems that an existing product aided design system adopts a single agent architecture and lacks an effective cooperation mechanism, so that an information island phenomenon is serious, and a unified product research and development scheme is difficult to form. According to the method, an assistant agent analyzes an issue and automatically calls cross-domain professional agents, each agent makes a proposal for the issue according to the own professional field, and the assistant agent organizes to perform review, screens out the proposal decided by the review, and arranges the proposal into a comprehensive decision scheme. The method is mainly used for optimizing the product research and development decision scheme.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a product development decision optimization method and system based on multi-agent systems. Background Technology

[0002] In the field of product development, existing artificial intelligence technologies are mainly manifested in product-aided design systems within a human-machine collaborative framework. These systems provide designers with design inspiration and preliminary solutions through AI modules, assisting them in completing complex product design tasks. Specifically, designers input design goals, performance indicators, functional requirements, and a series of design constraints into the AI ​​module, such as material properties, cost budgets, manufacturing process limitations, and dimensional boundaries. Based on preset algorithms and training data, the AI ​​module processes the input information, generating multiple possible design solutions or elements for designers to select, evaluate, and further optimize. In this model, AI technology primarily plays a supporting role; its output must be judged and decided by designers using their professional knowledge and experience to ultimately form a product design solution that meets the requirements.

[0003] However, existing product design-aided systems have significant limitations. These systems typically employ a single-agent architecture, providing suggestions only for specific domains and lacking multi-role collaborative decision-making capabilities. Decision-making processes among departments, such as R&D, operations, marketing, and finance, are relatively independent, lacking effective collaboration mechanisms. This results in severe information silos and makes it difficult to formulate unified product development plans. Summary of the Invention

[0004] To overcome the technical problem that existing product design-aided systems adopt a single agent architecture, lack effective collaboration mechanisms, resulting in severe information silos and difficulty in forming a unified product development plan, this invention provides a product development decision optimization method and system based on multiple agents.

[0005] This invention is achieved through the following technical solution:

[0006] Multi-agent-based product development decision optimization methods include:

[0007] S1: Receives one or more topics from user input, including R&D issues, initial solutions, and background information;

[0008] S2: The assistant agent analyzes the topic and, based on the analysis results, convenes at least three domain-specific agents to form a meeting group;

[0009] S3: Based on the analysis results, the assistant agent generates sub-tasks corresponding to the professional fields of each agent;

[0010] S4: Each domain-specific intelligent agent receives its own sub-task, accesses its own professional knowledge base, analyzes and calculates, and generates a preliminary proposal;

[0011] S5: The assistant agent receives preliminary proposals from all domain-specific agents, organizes one or more rounds of review and deliberation, and coordinates the interaction between agents to reach a consensus.

[0012] S6: The assistant agent generates and outputs the final decision-making scheme based on the consensus.

[0013] The assistant agent records the entire process data of this meeting and updates the professional knowledge base corresponding to the professional agents in each field based on the entire process data.

[0014] This invention uses an assistant agent to analyze a problem and automatically convenes cross-domain professional agents. Each agent makes a proposal based on its own professional field, and the assistant agent organizes a review and deliberation process to select the proposals that pass the review and compile them into a comprehensive decision-making solution. Each professional proposal has been verified by professional agents from other fields, ensuring that the decision-making solution derived by this invention has both depth across different professional fields and overall consistency.

[0015] This invention utilizes an assistant agent to record the entire process data of the meeting, and updates the professional knowledge bases of each domain's professional agents based on this data. When similar topics are discussed again, the domain agents can directly access the accumulated link data, significantly reducing the number of verification rounds and enabling subsequent decisions to maintain professional rigor while possessing self-evolution capabilities.

[0016] Furthermore, the specialized intelligent agent includes:

[0017] Develop an intelligent agent and configure it to perform technology assessments. Its professional knowledge base includes existing technology analysis data, technical difficulty data, and technology development roadmap templates.

[0018] The operational intelligence agent is configured to perform operational data analysis, and its professional knowledge base includes a user behavior pattern library, a market trend prediction model, and product feature priority ranking rules.

[0019] The marketing intelligence agent is configured to perform market analysis, and its professional knowledge base includes a competitor information database, a user segmentation profile database, and a marketing strategy case database.

[0020] The financial intelligence agent is configured to perform financial budget analysis. Its professional knowledge base includes cost-benefit calculation models, budget allocation scheme templates, and a library of indicators for evaluating the efficiency of fund use.

[0021] The risk intelligence agent is configured to perform risk assessments, and its professional knowledge base includes a risk feature library, a risk response strategy library, and a risk monitoring indicator library.

[0022] The new user intelligent agent is configured to perform user experience analysis, and its professional knowledge base includes a new user interaction scenario library, a user experience pain point knowledge library, and an interaction design optimization solution library.

[0023] The existing user intelligent agent is configured to perform user behavior analysis. Its professional knowledge base includes a user habit pattern library, personalized demand identification rules, and product improvement suggestion library.

[0024] This invention uses multi-domain specialized intelligent agents, each independently configured and deeply cultivating its own domain. Multiple intelligent agents can operate in parallel, are scalable, and are hot-swappable.

[0025] Furthermore, the review and deliberation process includes:

[0026] The assistant agent organizing meeting group votes on the preliminary proposals of all agents, marking the proposals that pass the vote as "adopted" and the proposals that fail the vote as "not adopted";

[0027] The assistant agent encapsulates the "unadopted" proposals and the corresponding reasons for the opposing votes into new topics, and organizes the meeting group to provide improvement suggestions for these new topics;

[0028] The assistant agent feeds back the improvement suggestions to the original proposal initiating agent, instructing the initiating agent to revise the original proposal or generate a new proposal.

[0029] The assistant agent organizes a meeting group to vote on revised or new proposals, forming an iterative optimization cycle;

[0030] After the assistant agent completes the iteration, it uses the proposals marked as "adopted" as the consensus.

[0031] If no consensus is reached after more than a preset number of iterations, all proposals associated with the "not adopted" proposal, the reasons for opposition, and suggestions for improvement will be packaged into a comprehensive report and submitted to the human administrator for final decision.

[0032] In the voting process, the proposing agent does not participate in voting on its own proposal.

[0033] This invention employs a review and deliberation process of "voting-opposing related factors-iterative re-proposal" to gradually converge and weaken biases in each proposal. Different agents simultaneously exert synergistic and constraining effects, preventing the formation of extremely optimal proposals in any particular domain. This ensures that the final consensus retains the professional depth of each domain while being globally acceptable, thus meeting the requirement of global uniformity.

[0034] Furthermore, a dynamic weighting mechanism is introduced into the generation of the voting results;

[0035] The dynamic weighting mechanism allows the assistant agent to dynamically adjust the voting weights of each participating agent based on the technical field of the current topic.

[0036] The higher the match between the agent's area of ​​expertise and the technical field of the current issue, the higher its voting weight will be.

[0037] This invention introduces a dynamic weighting mechanism, which tilts the weights toward the "more knowledgeable" intelligent agents, thereby enhancing the professionalism of the proposals.

[0038] Furthermore, the method also includes:

[0039] Based on the final decision-making scheme, the assistant agent generates an action plan that includes specific tasks, execution nodes, and acceptance criteria.

[0040] The assistant agent continuously tracks and monitors the progress and results of the action plan, and compares them with predetermined standards;

[0041] If a deviation is detected between the execution result and the predetermined standard, the assistant agent automatically triggers a new meeting process, encapsulates the deviation as a new topic, reconvenes domain-specific agents related to the deviation topic for one or more rounds of review and deliberation, and generates a corrective plan; the above steps are repeated until all tasks in the final decision plan are implemented in a closed loop.

[0042] This invention automatically transforms decision-making schemes into executable schemes through an assistant intelligence agent, avoiding misunderstandings and omissions that may occur when manually breaking down schemes, thus improving the efficiency of scheme execution. Through continuous tracking and monitoring mechanisms, execution deviations can be captured at the nascent stage and a review process can be triggered to prevent further expansion of deviations, thereby optimizing the sustainability of decision-making schemes throughout their entire lifecycle.

[0043] Another aspect of the present invention provides a product development decision optimization system based on multiple agents, including a data collection module, an assistant agent module, a multi-domain agent module, and a solution output module;

[0044] The data collection module is used to receive one or more topics input by the user, including research and development issues, initial solutions, and background information.

[0045] The assistant intelligent agent module includes a pre-meeting preparation component, a review and deliberation component, and a data management component;

[0046] The pre-meeting preparation component is used to parse the topic and, based on the parsing results, convene at least three domain-specific intelligent agents to form a meeting group; it is also used to generate sub-tasks corresponding to the professional domains of each domain-specific intelligent agent based on the parsing results.

[0047] The review and deliberation component is used to receive preliminary proposals from all domain-specific intelligent agents, organize one or more rounds of review and deliberation, and coordinate the interaction between the intelligent agents to reach a consensus.

[0048] The data management component is used to record the entire process data of this meeting and update the professional knowledge base corresponding to the professional intelligent agents in each field based on the entire process data.

[0049] The multi-domain intelligent agent module allows each domain-specific intelligent agent to receive its own sub-tasks, access its own professional knowledge base, and perform analysis and calculations to generate preliminary proposals.

[0050] The scheme output module is used to generate and output the final decision scheme based on the consensus.

[0051] Furthermore, the specialized intelligent agents in the multi-domain intelligent agent module include:

[0052] Develop an intelligent agent and configure it to perform technology assessments. Its professional knowledge base includes existing technology analysis data, technical difficulty data, and technology development roadmap templates.

[0053] The operational intelligence agent is configured to perform operational data analysis, and its professional knowledge base includes a user behavior pattern library, a market trend prediction model, and product feature priority ranking rules.

[0054] The marketing intelligence agent is configured to perform market analysis, and its professional knowledge base includes a competitor information database, a user segmentation profile database, and a marketing strategy case database.

[0055] The financial intelligence agent is configured to perform financial budget analysis. Its professional knowledge base includes cost-benefit calculation models, budget allocation scheme templates, and a library of indicators for evaluating the efficiency of fund use.

[0056] The risk intelligence agent is configured to perform risk assessments, and its professional knowledge base includes a risk feature library, a risk response strategy library, and a risk monitoring indicator library.

[0057] The new user intelligent agent is configured to perform user experience analysis, and its professional knowledge base includes a new user interaction scenario library, a user experience pain point knowledge library, and an interaction design optimization solution library.

[0058] The existing user intelligent agent is configured to perform user behavior analysis. Its professional knowledge base includes a user habit pattern library, personalized demand identification rules, and product improvement suggestion library.

[0059] Furthermore, the review and deliberation component is used to vote on the preliminary proposals of all agents, marking proposals that pass the vote as "adopted" and proposals that fail the vote as "not adopted";

[0060] It is also used to package "unadopted" proposals and the corresponding reasons for the opposing votes into new topics, and to organize the meeting group to give improvement suggestions on the new topics;

[0061] It is also used to feed back the improvement suggestions to the original proposal initiating agent, instructing the initiating agent to revise the original proposal or generate a new proposal;

[0062] It is also used to organize meeting groups to vote on revised or new proposals, forming an iterative optimization cycle;

[0063] It is also used to establish consensus by marking proposals as "adopted" after the loop iteration ends;

[0064] If no consensus is reached after more than a preset number of iterations, all proposals associated with the "not adopted" proposal, the reasons for opposition, and suggestions for improvement will be packaged into a comprehensive report and submitted to the human administrator for final decision.

[0065] In the voting process, the proposing agent does not participate in voting on its own proposal.

[0066] Furthermore, the review and deliberation component is configured to generate voting results using a dynamic weighting mechanism;

[0067] The dynamic weighting mechanism allows the assistant agent to dynamically adjust the voting weights of each participating agent based on the technical field of the current topic.

[0068] The higher the match between the agent's area of ​​expertise and the technical field of the current issue, the higher its voting weight will be.

[0069] Furthermore, the assistant intelligent agent module also includes a work follow-up component, which is used to generate an action plan containing specific tasks, execution nodes and acceptance criteria based on the final decision scheme;

[0070] It is also used to continuously track and monitor the progress and results of the action plan and compare them with predetermined standards;

[0071] If the monitoring detects a deviation between the execution result and the predetermined standard, the assistant agent automatically triggers a new meeting process, encapsulates the deviation as a new topic, and then reconvenes domain-specific agents related to the deviation topic for one or more rounds of review and deliberation to generate a corrective plan; the above steps are repeated until all tasks in the final decision plan are implemented in a closed loop.

[0072] The beneficial effects of this invention are:

[0073] Compared to the existing "single agent architecture," this invention uses an assistant agent to analyze the issue and automatically convene cross-domain professional agents. Each agent proposes a solution based on its own expertise, and the assistant agent selects the proposals that pass review and compiles them into a comprehensive decision-making scheme. The assistant agent records all data throughout the meeting and updates the professional knowledge bases of each professional agent based on this data. During the review process, agents from different professional fields submit their own proposals based on their expertise. Other professional agents vote on these proposals and provide suggestions for revising proposals that fail to pass. The agent that originally submitted the proposal then resubmits it or modifies the original proposal based on these suggestions. This process is organized by the assistant agent and completed collaboratively by agents from various professional fields. The final decision-making scheme integrates the suggestions of all agents, breaking down the information silos of the "single agent architecture" and avoiding data fragmentation.

[0074] Meanwhile, the assistant agent records the data throughout the entire process and updates each agent incrementally, achieving a closed loop where each agent becomes "smarter the more it is used"; the recording of the entire process data ensures a complete causal chain and supports audit backtracking. Attached Figure Description

[0075] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0076] Figure 1 This is a schematic diagram of one implementation process of the method of the present invention;

[0077] Figure 2 This is a schematic diagram of the structure of the system of the present invention;

[0078] Figure 3 This is a flowchart of the review and deliberation process in one embodiment of the present invention. Detailed Implementation

[0079] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific Implementation Method 1

[0081] refer to Figure 1 This embodiment demonstrates the implementation process of a product development decision optimization method based on multi-agent systems, specifically including the following steps:

[0082] S100: Receives one or more topics from user input, including R&D issues, initial solutions, and background information.

[0083] The R&D issues include records of technical challenges, market demand data, and user experience defect data. Collecting these R&D issues ensures that the decision-making objectives for optimization solutions correspond one-to-one with the real pain points, preventing the optimization direction from deviating.

[0084] The initial solution is a solution proposed by the R&D team based on existing information to address the R&D problem, including product functions, design, and technical roadmap.

[0085] The background information includes market analysis, user feedback, and competitor analysis.

[0086] By collecting research and development problems, initial solutions, and background information, all subsequent agents can reason based on the same complete set of facts, eliminating decision-making conflicts caused by information silos.

[0087] S200: The assistant agent analyzes the topic and, based on the analysis results, convenes at least three domain-specific agents to form a conference group.

[0088] The specialized intelligent agents include:

[0089] Develop an intelligent agent and configure it to perform technology assessments. Its professional knowledge base includes existing technology analysis data, technical difficulty data, and technology development roadmap templates.

[0090] The operational intelligence agent is configured to perform operational data analysis, and its professional knowledge base includes a user behavior pattern library, a market trend prediction model, and product feature priority ranking rules.

[0091] The marketing intelligence agent is configured to perform market analysis, and its professional knowledge base includes a competitor information database, a user segmentation profile database, and a marketing strategy case database.

[0092] The financial intelligence agent is configured to perform financial budget analysis. Its professional knowledge base includes cost-benefit calculation models, budget allocation scheme templates, and a library of indicators for evaluating the efficiency of fund use.

[0093] The risk intelligence agent is configured to perform risk assessments, and its professional knowledge base includes a risk feature library, a risk response strategy library, and a risk monitoring indicator library.

[0094] The new user intelligent agent is configured to perform user experience analysis, and its professional knowledge base includes a new user interaction scenario library, a user experience pain point knowledge library, and an interaction design optimization solution library.

[0095] The existing user intelligent agent is configured to perform user behavior analysis. Its professional knowledge base includes a user habit pattern library, personalized demand identification rules, and product improvement suggestion library.

[0096] S300: Based on the analysis results, the assistant agent generates sub-tasks corresponding to the professional fields of each agent.

[0097] S400: Specialized intelligent agents in various fields receive their respective sub-tasks, access their respective professional knowledge bases, analyze and calculate, and generate preliminary proposals.

[0098] The proposal process for the specialized intelligent agents in each field includes:

[0099] The research and development of intelligent agents is based on the corresponding professional knowledge base data and the received sub-tasks. They conduct in-depth analysis of the technical feasibility of the product, assess the project development cycle, and propose architecture optimization proposals.

[0100] Based on the corresponding professional knowledge base data and the received sub-tasks, the operational intelligence agent proposes suggestions for adjusting the function priority based on user retention rate and activity indicators, and provides a design plan for canary release.

[0101] Based on relevant professional knowledge base data and received sub-tasks, the marketing intelligence agent estimates customer acquisition costs for each channel, optimizes product selling point packaging strategies, and proposes phased promotion plans.

[0102] Based on the corresponding professional knowledge base data and the received sub-tasks, the financial intelligent agent conducts sensitivity analysis on various budget allocations and proposes cost control proposals from the perspective of fund utilization efficiency.

[0103] Based on the corresponding professional knowledge base data and the received sub-tasks, the risk intelligence agent identifies potential compliance risk points, improves user agreement terms, establishes a public opinion mechanism, and proposes data security protection plans.

[0104] Based on the corresponding professional knowledge base data and the received sub-tasks, the new user intelligent agent simulates the scenario of the user's first use, focuses on the pain points of user experience, and proposes optimization proposals for the registration process and the new user guidance process.

[0105] Based on relevant professional knowledge base data and received sub-tasks, the existing user intelligent agent proposes suggestions that include personalized usage paths and optimizations to the membership growth system.

[0106] S500: The assistant agent receives preliminary proposals from all domain-specific agents, organizes one or more rounds of review and deliberation, and coordinates the interactions between agents to reach a consensus.

[0107] refer to Figure 3 The review and deliberation process includes:

[0108] The assistant agent organizing meeting group votes on the preliminary proposals of all agents, marking the proposals that pass the vote as "adopted" and the proposals that fail the vote as "not adopted";

[0109] The assistant agent encapsulates the "unadopted" proposals and the corresponding reasons for the opposing votes into new topics, and organizes the meeting group to provide improvement suggestions for these new topics;

[0110] The assistant agent feeds back the improvement suggestions to the original proposal initiating agent, instructing the initiating agent to revise the original proposal or generate a new proposal.

[0111] The assistant agent organizes a meeting group to vote on revised or new proposals, forming an iterative optimization cycle;

[0112] If no consensus is reached after more than a preset number of iterations, all proposals associated with the "not adopted" proposal, the reasons for opposition, and suggestions for improvement will be packaged into a comprehensive report and submitted to the human administrator for final decision.

[0113] In the voting process, the proposing agent does not participate in voting on its own proposal.

[0114] The voting process can be conducted anonymously.

[0115] The generation of the voting results incorporates a dynamic weighting mechanism;

[0116] The dynamic weighting mechanism allows the assistant agent to dynamically adjust the voting weights of each participating agent based on the technical field of the current topic.

[0117] The higher the match between the agent's area of ​​expertise and the technical field of the current issue, the higher its voting weight will be.

[0118] S600: Generate and output the final decision scheme based on the consensus.

[0119] The assistant agent records the entire process data of this meeting and updates the professional knowledge base corresponding to the professional agents in each field based on the entire process data. Specific Implementation Method Two

[0121] refer to Figure 2This embodiment demonstrates a product development decision optimization system based on multi-agent intelligence, including: a data collection module, an assistant intelligence module, a multi-domain intelligence module, and a solution output module.

[0122] The data collection module is used to receive one or more topics input by the user, including research and development issues, initial solutions, and background information.

[0123] The assistant intelligent agent module includes a pre-meeting preparation component, a review and deliberation component, and a data management component;

[0124] The pre-meeting preparation component is used to parse the topic and, based on the parsing results, convene at least three domain-specific intelligent agents to form a meeting group; it is also used to generate sub-tasks corresponding to the professional domains of each domain-specific intelligent agent based on the parsing results.

[0125] The review and deliberation component is used to receive preliminary proposals from all domain-specific intelligent agents, organize one or more rounds of review and deliberation, and coordinate the interaction between the intelligent agents to reach a consensus.

[0126] The data management component is used to record the entire process data of this meeting and update the professional knowledge base corresponding to the professional intelligent agents in each field based on the entire process data.

[0127] The multi-domain intelligent agent module allows each domain-specific intelligent agent to receive its own sub-tasks, access its own professional knowledge base, and perform analysis and calculations to generate preliminary proposals.

[0128] The scheme output module is used to generate and output the final decision scheme based on the consensus.

[0129] The specialized intelligent agents in the multi-domain intelligent agent module include:

[0130] Develop an intelligent agent and configure it to perform technology assessments. Its professional knowledge base includes existing technology analysis data, technical difficulty data, and technology development roadmap templates.

[0131] The operational intelligence agent is configured to perform operational data analysis, and its professional knowledge base includes a user behavior pattern library, a market trend prediction model, and product feature priority ranking rules.

[0132] The marketing intelligence agent is configured to perform market analysis, and its professional knowledge base includes a competitor information database, a user segmentation profile database, and a marketing strategy case database.

[0133] The financial intelligence agent is configured to perform financial budget analysis. Its professional knowledge base includes cost-benefit calculation models, budget allocation scheme templates, and a library of indicators for evaluating the efficiency of fund use.

[0134] The risk intelligence agent is configured to perform risk assessments, and its professional knowledge base includes a risk feature library, a risk response strategy library, and a risk monitoring indicator library.

[0135] The new user intelligent agent is configured to perform user experience analysis, and its professional knowledge base includes a new user interaction scenario library, a user experience pain point knowledge library, and an interaction design optimization solution library.

[0136] The existing user intelligent agent is configured to perform user behavior analysis. Its professional knowledge base includes a user habit pattern library, personalized demand identification rules, and product improvement suggestion library.

[0137] The review and deliberation component is used to vote on the preliminary proposals of all agents, marking proposals that pass the vote as "adopted" and proposals that fail the vote as "not adopted".

[0138] It is also used to package "unadopted" proposals and the corresponding reasons for the opposing votes into new topics, and to organize the meeting group to give improvement suggestions on the new topics;

[0139] It is also used to feed back the improvement suggestions to the original proposal initiating agent, instructing the initiating agent to revise the original proposal or generate a new proposal;

[0140] It is also used to organize meeting groups to vote on revised or new proposals, forming an iterative optimization cycle;

[0141] It is also used to establish consensus by marking proposals as "adopted" after the loop iteration ends;

[0142] If no consensus is reached after more than a preset number of iterations, all proposals associated with the "not adopted" proposal, the reasons for opposition, and suggestions for improvement will be packaged into a comprehensive report and submitted to the human administrator for final decision.

[0143] In the voting process, the proposing agent does not participate in voting on its own proposal.

[0144] The review and deliberation component is configured to generate voting results using a dynamic weighting mechanism;

[0145] The dynamic weighting mechanism allows the assistant agent to dynamically adjust the voting weights of each participating agent based on the technical field of the current topic.

[0146] The higher the match between the agent's area of ​​expertise and the technical field of the current issue, the higher its voting weight will be.

[0147] The assistant intelligent agent module also includes a work follow-up component, which is used to generate an action plan containing specific tasks, execution nodes and acceptance criteria based on the final decision scheme.

[0148] It is also used to continuously track and monitor the progress and results of the action plan and compare them with predetermined standards;

[0149] If the monitoring detects a deviation between the execution result and the predetermined standard, the assistant agent automatically triggers a new meeting process, encapsulates the deviation as a new topic, and then reconvenes domain-specific agents related to the deviation topic for one or more rounds of review and deliberation to generate a corrective plan; the above steps are repeated until all tasks in the final decision plan are implemented in a closed loop.

[0150] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0151] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A product development decision optimization method based on multi-agent systems, characterized in that, include: S1: Receives one or more topics from user input, including R&D issues, initial solutions, and background information; S2: The assistant agent analyzes the topic and, based on the analysis results, convenes at least three domain-specific agents to form a meeting group; S3: Based on the analysis results, the assistant agent generates sub-tasks corresponding to the professional fields of each agent; S4: Each domain-specific intelligent agent receives its own sub-task, accesses its own professional knowledge base, analyzes and calculates, and generates a preliminary proposal; S5: The assistant agent receives preliminary proposals from all domain-specific agents, organizes one or more rounds of review and deliberation, and coordinates the interaction between agents to reach a consensus. S6: The assistant agent generates and outputs the final decision scheme based on the consensus. The assistant agent records the entire process data of this meeting and updates the professional knowledge base corresponding to the professional agents in each field based on the entire process data.

2. The product development decision optimization method based on multi-agent systems according to claim 1, characterized in that, The specialized intelligent agents include: Develop an intelligent agent and configure it to perform technology assessments. Its professional knowledge base includes existing technology analysis data, technical difficulty data, and technology development roadmap templates. The operational intelligence agent is configured to perform operational data analysis, and its professional knowledge base includes a user behavior pattern library, a market trend prediction model, and product feature priority ranking rules. The marketing intelligence agent is configured to perform market analysis, and its professional knowledge base includes a competitor information database, a user segmentation profile database, and a marketing strategy case database. The financial intelligence agent is configured to perform financial budget analysis. Its professional knowledge base includes cost-benefit calculation models, budget allocation scheme templates, and a library of indicators for evaluating the efficiency of fund use. The risk intelligence agent is configured to perform risk assessments, and its professional knowledge base includes a risk feature library, a risk response strategy library, and a risk monitoring indicator library. The new user intelligent agent is configured to perform user experience analysis, and its professional knowledge base includes a new user interaction scenario library, a user experience pain point knowledge library, and an interaction design optimization solution library. The existing user intelligent agent is configured to perform user behavior analysis. Its professional knowledge base includes a user habit pattern library, personalized demand identification rules, and product improvement suggestion library.

3. The product development decision optimization method based on multi-agent systems according to claim 1, characterized in that, The review and deliberation process includes: The assistant agent organizing meeting group votes on the preliminary proposals of all agents, marking the proposals that pass the vote as "adopted" and the proposals that fail the vote as "not adopted"; The assistant agent encapsulates the "not adopted" proposals and the corresponding reasons for the opposing votes into new topics, and organizes the meeting group to provide improvement suggestions on the new topics; The assistant agent feeds back the improvement suggestions to the original proposal initiating agent, instructing the initiating agent to revise the original proposal or generate a new proposal. The assistant agent organizes a meeting group to vote on revised or new proposals, forming an iterative optimization cycle; After the loop iteration is completed, the assistant agent will use the proposals marked as "adopted" as the consensus; If no consensus is reached after more than a preset number of iterations, all proposals associated with the "not adopted" proposal, the reasons for opposition, and suggestions for improvement will be packaged into a comprehensive report and submitted to the human administrator for final decision. In the voting process, the proposing agent does not participate in voting on its own proposal.

4. The product development decision optimization method based on multi-agent systems according to claim 3, characterized in that, The generation of the voting results incorporates a dynamic weighting mechanism; The dynamic weighting mechanism allows the assistant agent to dynamically adjust the voting weights of each participating agent based on the technical field of the current topic. The higher the match between the agent's area of ​​expertise and the technical field of the current issue, the higher its voting weight will be.

5. The product development decision optimization method based on multi-agent systems according to claim 1, characterized in that, The method further includes: Based on the final decision-making scheme, the assistant agent generates an action plan that includes specific tasks, execution nodes, and acceptance criteria. The assistant agent continuously tracks and monitors the progress and results of the action plan, and compares them with predetermined standards; If a deviation is detected between the execution result and the predetermined standard, the assistant agent automatically triggers a new meeting process, encapsulates the deviation as a new topic, reconvenes domain-specific agents related to the deviation topic for one or more rounds of review and deliberation, and generates a corrective plan; the above steps are repeated until all tasks in the final decision plan are implemented in a closed loop.

6. A product development decision optimization system based on multi-agent systems, characterized in that, It includes a data collection module, an assistant intelligent agent modality, a multi-domain intelligent agent module, and a solution output module; The data collection module is used to receive one or more topics input by the user, including research and development issues, initial solutions, and background information. The assistant intelligent agent module includes a pre-meeting preparation component, a review and deliberation component, and a data management component; The pre-meeting preparation component is used to analyze the topic and, based on the analysis results, convene at least three domain-specific intelligent agents to form a meeting group. It is also used to generate sub-tasks corresponding to the professional fields of professional intelligent agents in each field based on the analysis results; The review and deliberation component is used to receive preliminary proposals from all domain-specific intelligent agents, organize one or more rounds of review and deliberation, and coordinate the interaction between the intelligent agents to reach a consensus. The data management component is used to record the entire process data of this meeting and update the professional knowledge base corresponding to the professional intelligent agents in each field based on the entire process data. The multi-domain intelligent agent module allows each domain-specific intelligent agent to receive its own sub-tasks, access its own professional knowledge base, analyze and calculate, and generate preliminary proposals. The scheme output module is used to generate and output the final decision scheme based on the consensus.

7. The product development decision optimization system based on multi-agent technology according to claim 6, characterized in that, The specialized intelligent agents in the multi-domain intelligent agent module include: Develop an intelligent agent and configure it to perform technology assessments. Its professional knowledge base includes existing technology analysis data, technical difficulty data, and technology development roadmap templates. The operational intelligence agent is configured to perform operational data analysis, and its professional knowledge base includes a user behavior pattern library, a market trend prediction model, and product feature priority ranking rules. The marketing intelligence agent is configured to perform market analysis, and its professional knowledge base includes a competitor information database, a user segmentation profile database, and a marketing strategy case database. The financial intelligence agent is configured to perform financial budget analysis. Its professional knowledge base includes cost-benefit calculation models, budget allocation scheme templates, and a library of indicators for evaluating the efficiency of fund use. The risk intelligence agent is configured to perform risk assessments, and its professional knowledge base includes a risk feature library, a risk response strategy library, and a risk monitoring indicator library. The new user intelligent agent is configured to perform user experience analysis, and its professional knowledge base includes a new user interaction scenario library, a user experience pain point knowledge library, and an interaction design optimization solution library. The existing user intelligent agent is configured to perform user behavior analysis. Its professional knowledge base includes a user habit pattern library, personalized demand identification rules, and product improvement suggestion library.

8. The product development decision optimization system based on multi-agent technology according to claim 6, characterized in that, The review and deliberation component is used to vote on the preliminary proposals of all agents, marking proposals that pass the vote as "adopted" and proposals that fail the vote as "not adopted". It is also used to package "unadopted" proposals and the corresponding reasons for the opposing votes into new topics, and to organize the meeting group to give improvement suggestions on the new topics; It is also used to feed back the improvement suggestions to the original proposal initiating agent, instructing the initiating agent to revise the original proposal or generate a new proposal; It is also used to organize meeting groups to vote on revised or new proposals, forming an iterative optimization cycle; It is also used to establish consensus by marking proposals as "adopted" after the loop iteration ends; If no consensus is reached after more than a preset number of iterations, all proposals associated with the "not adopted" proposal, the reasons for opposition, and suggestions for improvement will be packaged into a comprehensive report and submitted to the human administrator for final decision. In the voting process, the proposing agent does not participate in voting on its own proposal.

9. The product development decision optimization system based on multi-agent technology according to claim 8, characterized in that, The review and deliberation component is configured to generate voting results using a dynamic weighting mechanism; The dynamic weighting mechanism allows the assistant agent to dynamically adjust the voting weights of each participating agent based on the technical field of the current topic. The higher the match between the agent's area of ​​expertise and the technical field of the current issue, the higher its voting weight will be.

10. The product development decision optimization system based on multi-agent technology according to claim 6, characterized in that, The assistant intelligent agent module also includes a work follow-up component, which is used to generate an action plan containing specific tasks, execution nodes and acceptance criteria based on the final decision scheme. It is also used to continuously track and monitor the progress and results of the action plan and compare them with predetermined standards; If a deviation is detected between the execution result and the predetermined standard, the assistant agent automatically triggers a new meeting process, encapsulates the deviation as a new topic, and then reconvenes domain-specific agents related to the deviation topic for one or more rounds of review and deliberation to generate a corrective plan. The above steps are repeated until all tasks in the final decision plan are implemented in a closed loop.

Images