Agent-based concept innovation design system and implementation method

Abstract

An agent-based concept innovation design system and implementation method. A design process is regulated using a large language model, so that the design process is intelligent, thereby greatly improving product concept design efficiency. A design template built from an existing design process is input into the large language model, so that the large language model can regulate the design process according to previous design experience, thereby using existing design experience to perform product concept design to obtain a concept innovation design product required by a user.

Description

An agent-based concept innovation design system and implementation method TECHNICAL FIELD

[0001] The present application belongs to the field of product design software, and specifically relates to an agent-based concept innovation design system and implementation method. BACKGROUND

[0002] In the product development process, concept design is the key link to transform innovative ideas into actual products. It not only requires designers to have a deep understanding of market demand and technical feasibility, but also needs to make innovations and trade-offs in terms of function, appearance, material selection, etc. This process inevitably involves the integration of cross-disciplinary knowledge and the extensive use of innovative thinking.

[0003] With the development of generative artificial intelligence (AIGC) technology, large language models have shown potential in the field of concept design. For example, using large language models to analyze and organize user requirements, using image generation models to quickly generate concept sketches, etc. These applications provide effective support for various aspects of concept design. Large language model technology can use its data analysis, prototype drawing, and scheme evaluation capabilities to significantly improve the efficiency of the design process and the quality of the design results.

[0004] The invention patent application with publication number CN116955613A discloses a method for generating product concepts based on research data and large language models, which includes: obtaining the theme, source and publication time of the research data, classifying the research content, constructing an industry knowledge base, and vectorizing the knowledge base; extracting research knowledge from the industry knowledge base, and classifying product concepts that need to be generated; generating prompts for product attributes under each product feature classification according to the functional characteristics of the product to be generated; according to the use habits, skills and knowledge background of product designers, automatically recommend structured parameters of prompts suitable for different products according to the functional characteristics of different products; according to different user needs, adjust the structured parameters in the parameter list to generate prompt content that accurately describes the product. However, this patent application requires manual intervention for analysis and lacks intelligence and automation.

[0005] In the concept design stage, designers usually need to invest a lot of time in market research, collecting relevant design cases, and making prototype concept design drawings and three-dimensional models, and also need to make trade-offs among numerous design schemes. However, there is no report in the prior art that applies large language models to the design process that requires a lot of effort to help designers efficiently complete product concept design.

[0006] Therefore, it is urgent to design an agent based on a large language model, a concept innovation design system assisted by a large language model in different stages of requirement analysis, function design, logic design, and prototype design, so as to realize intelligentization and automation of concept design. SUMMARY

[0007] The application provides an agent-based concept innovation design system, which can help designers efficiently complete concept innovation design and obtain design results needed by designers.

[0008] The embodiment of the application provides a concept innovation design system based on an agent, which comprises: a concept innovation design module comprising a design flow graphical unit and a large language model dialogue unit, the design flow graphical unit is used for patterning design assets and innovation methods to obtain a first design flow, the large language model dialogue unit is used for interacting with a user through a large language model, and a new design flow is obtained by regulating the first design flow according to a process order of a design template based on information input by the user, the new design flow is stored in a design flow creation unit, and a final design asset, that is, a design result, is obtained through the new design flow; a design template management module used for generating a design template and sending the design template to the large language model; an innovation method module used for storing an innovation method and calling the innovation method based on information input by the user, and outputting a new design asset through the innovation method; and a design asset management module used for storing design assets and searching for design assets suitable for information input by the user.

[0009] Preferably, the large language model dialogue unit is used for obtaining a new design flow by regulating the first design flow according to a process order of a design template based on information input by the user, wherein the regulation of the first design flow comprises one or more of adding a new design asset in the first design flow, adding an innovation method, adding an edge to an existing design asset, adjusting a parameter of a function corresponding to the innovation method, and running a set innovation method; and the information input by the user comprises text description input by the user, design assets, and innovation methods.

[0010] Preferably, the new design flow is obtained by regulating the first design flow according to the process order of the design template, which comprises that the large language model interacts with the user in real time through the process order of the design template, and regulates the first design flow in real time based on information obtained in the interaction with the user.

[0011] Preferably, the design flow graphing unit is configured to graphically pattern the design assets and the innovation method to obtain a first design flow, wherein the innovation method and the design asset in the first design flow are taken as nodes, if the design asset is an input of the innovation method, a directed edge from the design asset to the innovation method is set, if the design asset is an output of the innovation method, a directed edge from the innovation method to the design asset is set, and the design asset has one directed edge from the innovation method.

[0012] Preferably, the conceptual innovation design module further comprises a design flow management unit configured to create, open, edit the name, delete and copy the design flow.

[0013] Preferably, the conceptual innovation design module further comprises a design asset traceability unit configured to save the corresponding design flow when a new design asset is generated, and convert the corresponding design flow into a graph for display to the user.

[0014] Preferably, the innovation method module comprises an innovation method display unit and an innovation method calling unit; the innovation method display unit is configured to search for an innovation method based on a user purpose instruction, the innovation method comprises an innovation method type and an innovation method parameter, search for a corresponding innovation method type based on the user purpose instruction, display the innovation method parameter corresponding to the innovation method type to the user through a large language model, and complete the setting of the innovation method parameter based on the selection instruction of the user on the parameter received through the large language model, so as to obtain an innovation method adapted to the user purpose instruction; the innovation method calling unit is configured to obtain a new design asset by running the adapted innovation method based on the design asset obtained from the design asset management module and adapted to the information input by the user.

[0015] Preferably, the design asset management module comprises a multi-modal design asset storage unit, a cross-modal asset searching unit, a design asset management unit and a design asset import and export unit; the multi-modal design asset storage unit is configured to save the design assets of different modalities in the form of binary data; the cross-modal asset searching unit is configured to convert the multi-modal design assets into vectors to obtain a vector database, convert the information input by the user into a user vector in the searching process, match the user vector with the vectors in the vector database, obtain vectors with high matching degrees and corresponding ids, and find the design assets adapted to the information input by the user based on the ids; the design asset management unit is configured to copy, delete, edit the name and view the design assets; and the design asset import and export unit is configured to import the design assets from external devices or download the design assets to external devices.

[0016] Preferably, the design template management module comprises a design template generation unit and a design template display unit; the design template generation unit is configured to input a first design process into a large language model to obtain a corresponding text description, construct a design template through the first design process and the corresponding text description, and send the design template to the large language model; and the design template display unit is configured to graphically represent the first design process, represent the text description in a text form, and display the graphically represented first design process and the text description in the text form to a user through the large language model.

[0017] In another aspect, a concept innovation design method based on an agent, which obtains a design result based on input information of a user by using the concept innovation design system based on the agent, comprises: receiving information input by the user in real time by using the large language model, regulating an existing design process to obtain a new design process according to a flow sequence of a design template based on the input information of the user, and obtaining a final design asset of the new design process, i.e., a design result, to complete the concept innovation design.

[0018] Compared with the prior art, the present application has the following beneficial effects: the present application regulates a design process by using a large language model, thereby making the design process intelligent and greatly improving the efficiency of generating a concept design; and the present application inputs a design template constructed by an existing design process into a large language model, so that the large language model can regulate a design process according to previous design experience, thereby enabling the product concept design to utilize the existing design experience to obtain a concept innovation design product required by a user. BRIEF DESCRIPTION OF DRAWINGS

[0019] Fig. 1 is a structural composition block diagram of a concept innovation design system based on an agent according to an embodiment of the present application; Fig. 2 is a block diagram of a concept innovation design system based on an agent according to an embodiment of the present application; Fig. 3 is an interface layout diagram of a concept innovation design system based on an agent according to an embodiment of the present application; Fig. 4 is an interface diagram of a directory area according to an embodiment of the present application; Fig. 5 is an analysis interface diagram according to an embodiment of the present application; and Fig. 6 is an interactive interface diagram of a planner and a designer according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions of the present application will be described below in detail with reference to the accompanying drawings and embodiments, but the present application is not limited in any way by the embodiments.

[0021] In order to help designers to efficiently obtain a conceptual innovative design meeting the requirements of the designers, the embodiment of the present application provides an agent-based conceptual innovative design system which interacts with the user by using the innovative design template, such as informing the user whether market research, user research, etc. is needed, and regulates the existing design process to obtain a new design process, and obtains the final design result based on the new design process.

[0022] The embodiment of the present application provides an agent-based conceptual innovative design system, as shown in FIG. 1 and FIG. 2, which comprises a conceptual innovative design module, a design asset management module, a design template management module and an innovative method module.

[0023] The embodiment of the present application provides a conceptual innovative design module which comprises a design process graphical unit and a large language model dialogue unit, wherein the design process graphical unit is used to represent the design assets and the innovative methods as nodes, and the relationship between the design assets and the innovative methods as directed edges, specifically, if an innovative method Mi uses a design asset Ai as input, the design asset Ai has a directed edge pointing to the innovative method Mi; if an innovative method Mi produces a design asset Ao as output, the innovative method Mi has a directed edge pointing to the design asset Ao, wherein a specific design asset can only have one directed edge pointing to it, but the design asset can have multiple directed edges pointing to different innovative methods.

[0024] In an embodiment, the design process interface provided by the embodiment supports certain operations on the design assets and the innovative methods in a graphical manner, including the following contents: (1) adding a new design asset in the design process.

[0025] (2) adding a new innovative method in the design process.

[0026] (3) adding a directed edge for the existing design asset to connect it to the innovative method, so that the innovative method can accept the design asset as input.

[0027] (4) adjusting the parameters of the existing innovative method.

[0028] (5) running a specific innovative method.

[0029] The large language model dialogue unit provided by the embodiment of the present application is used for patterning design assets and innovative methods to obtain a first design process, that is, an existing design process, and is used for interacting with a user through a large language model, and regulating a first design process according to a process order of a design template based on information input by the user to obtain a new design process. The large language model adopted by the embodiment of the present application is an existing large language model. In an embodiment, the large language model adopted by the embodiment is Zhi Puxing glm-4.

[0030] Specifically, the user can dialogue with the large language model in the form of pure text, and can also quote design assets and innovative methods in the dialogue process. When the user inputs the design assets or the innovative methods, the design template management module generates the description information of the design assets or the innovative methods, and embeds the description information of the innovative methods into the dialogue input to help the large language model understand the input of the user.

[0031] The large language model implements operations on the design assets and the innovative methods. When replying to the user, the large language model can perform certain operations in addition to pure text replies, including the following contents: (adjusting the design process through the input of the user to obtain at least one required result) (1) querying corresponding design assets from a design asset library.

[0032] (2) adding a new design asset in the design process.

[0033] (3) adding a new innovative method in the design process.

[0034] (4) adding a directed edge to the existing design asset, and connecting the existing design asset to the innovative method, so that the innovative method can accept the design asset as input.

[0035] (5) adjusting parameters of the existing innovative method.

[0036] (6) running a specific innovative method.

[0037] The concept innovation design module provided by the embodiment of the present application further includes a design asset traceability unit. The design asset traceability unit is used for saving a corresponding design process when a new design asset is generated by an innovative method, and converting the corresponding design process of the new design asset into a graph to show the user.

[0038] The concept innovation design module provided by the embodiment of the present application further includes a design process management unit. The design process management unit is used for new creation, opening, editing of a name, deletion and copy operation of a design process.

[0039] The design template management module provided by the embodiment of the present application is used for generating a design template and sending the design template to a large language model.

[0040] Specifically, the design template management module provided by the embodiment of the present application comprises a design template generation unit and a design template display unit, wherein the design template generation unit is used for inputting a first design process into a large language model to obtain a corresponding text description, constructing a design template through the first design process and the corresponding text description, and sending the design template to the large language model, so that the large language model designs according to the steps corresponding to the design template.

[0041] The design template display unit provided by the embodiment of the present application is used for graphically representing the first design process and representing the text description in a text form, and displaying the graphically represented first design process and the text description in the text form to a user through the large language model.

[0042] The innovation method module provided by the embodiment of the present application is used for storing an innovation method, calling the innovation method based on a user purpose instruction output by the large language model, outputting a new design asset through the innovation method, converting information input by the user into the user purpose instruction through the large language model, and sending the user purpose instruction to the innovation method module.

[0043] The innovation method module provided by the embodiment of the present application comprises an innovation method display unit and an innovation method calling unit, the innovation method display unit is used for searching an innovation method based on a user purpose instruction, the innovation method comprises an innovation method type and an innovation method parameter, a corresponding innovation method type is searched based on the user purpose instruction, a corresponding function is searched based on the innovation method type, in a specific embodiment, the function comprises an image generation model driven creative divergence such as Stable Diffusion, a large language model driven TRIZ design method, a FBS analysis driven by a big prophecy model, etc., a corresponding innovation method is obtained, the function parameters corresponding to the innovation method are displayed to the user through the large language model, and the function parameters corresponding to the innovation method are set based on a selection instruction of the user on the parameters received through the large language model, so that an innovation method adapted to the user purpose instruction is obtained.

[0044] The innovation method calling unit provided by the embodiment of the present application is used for obtaining a new design asset by running the adapted innovation method based on the input design asset obtained from the design asset library.

[0045] The innovative method provided by the embodiment of the application is a design method used in a design process and a corresponding artificial intelligence method, and specifically includes creative divergence driven by an image generation model such as Stable Diffusion and a TRIZ design method driven by a large language model.

[0046] The asset management module provided by the embodiment of the application includes a multi-modal design asset storage unit, a cross-modal asset search unit, a design asset management unit, and a design asset import and export unit.

[0047] The multi-modal design asset storage unit provided by the embodiment of the application is used to save design assets of different modalities in the form of binary data.

[0048] The cross-modal asset search unit provided by the embodiment of the application is used to convert multi-modal design assets into vectors to obtain a vector database, convert user input information into a user vector in a search process, match the user vector with vectors in the vector database, obtain vectors with high matching degrees and corresponding ids, and find design assets based on the ids as input design assets.

[0049] Specifically, converting multi-modal design assets into vectors to obtain a vector database includes: directly representing assets of text, image, and audio types into vectors based on a multi-modal model; summarizing a video into text based on the multi-modal model and representing the text into a vector through the aforementioned multi-modal model for text; mapping 3D model assets into three views, combining to form an image, and representing the image into a vector through the aforementioned multi-modal model for image; for assets of neural network models and program codes, extracting artificially set description information to form text and representing the text into a vector through the aforementioned multi-modal model for text. Finally, the generated vector data is stored in the vector database.

[0050] The design asset management unit provided by the embodiment of the application is used for copying, deleting, name editing, and viewing design assets; and the design asset import and export unit provided by the embodiment of the application is used for importing design assets from external devices or downloading design assets to external devices.

[0051] Specifically, the embodiment of the application uploads a design asset file from a computer locally, and a user specifies the name, type, and description of the asset. The system saves it to a database. The embodiment of the application downloads the design asset in the form of a file.

[0052] The design assets provided by the embodiments of the present application are a series of types of binary files readable by a system, including text, image, audio, video, 3D model, neural network and the like. The design assets are reusable design elements, and are input and output forms of the innovation method.

[0053] The design assets stored by the design asset management module and the innovation methods stored by the innovation method module can be added to the interface through graphical operations and displayed in the form of a graphical design process.

[0054] In another aspect, the present application also provides an agent-based concept innovation design method, characterized in that the agent-based concept innovation design system is used to obtain a design result based on user input information, including: using the large language model to receive user input information in real time, based on the user input information, regulating the existing design process according to the process order of the design template to obtain a new design process, and based on the final design asset of the new design process, i.e., the design result, to complete the concept innovation design.

[0055] In a specific embodiment, the present embodiment provides a method for helping designers to realize concept innovation design using the agent-based concept innovation design system, and the specific steps are as follows: 1. Selecting an FBS analysis template in the design template (design template management module) interface to newly create a design process, i.e., an FBS process.

[0056] 2. The designer Teddy inputs "I want to design an L-shaped, cloth material, comfortable sofa" in the large language model dialogue unit, and the large language model used in the present embodiment is ZhiPuxing GLM-4.

[0057] 3. The large model responds and outputs: According to the FBS process, the design of "L-shaped, cloth, comfortable sofa" will be carried out, and the following steps will be carried out: FBS requirement analysis (such as requirement analysis-method 1-vywq in FIG. 5): The requirement input by the designer Teddy is automatically summarized as design constraints, and based on the design constraints, the basic functional requirements that need to be met by the design are determined, such as comfort, material, shape (L-shaped) and the like. According to your description, I will generate a text asset describing the requirements.

[0058] FBS function analysis: Based on the pre-analysis, i.e., the text asset describing the requirements generated by the FBS requirement analysis, the functions that the sofa must implement are analyzed, which may include its support capability, suitable sitting posture support, easy cleaning and durability and the like.

[0059] FBS behavior analysis: Based on the results of the function analysis, the behavior patterns of the user interacting with the sofa are further analyzed, such as sitting down, lying down, getting up and the like.

[0060] FBS Structure Analysis: Analyze the specific structure of the sofa design to support the expected behavior, including frame design, fabric selection, filling materials, etc.

[0061] FBS Emotional Analysis: Consider how the design should evoke emotional resonance in users, including color, texture, and overall design aesthetics.

[0062] FBS Image Generation and Model Making: Based on the results of structure and emotional analysis, generate images and 3D models of the sofa for further evaluation and modification.

[0063] Next, I will create a text asset for you that details the requirements for the sofa design. Are you ready? You can provide more details or directions to accurately describe these requirements.

[0064] 4. User input: "Please proceed with the FBS process for design."

[0065] 5. The large language model creates a design asset that describes the user's requirements and asks if the user wants to continue with the above template.

[0066] 6. The user inputs "yes," and the large language model adds the FBS requirement analysis method to the left canvas.

[0067] 7. The user clicks the "run" button of the left method, runs the method, and generates requirement information.

[0068] 8. The user continues to input "yes," and the large model adds each method to the left canvas in turn, while the user runs each method.

[0069] 9. After the user completes the design according to the process, the large language model responds: If you have other design requirements or need to perform other operations, such as evaluation, modification, or further design iteration, please let me know. If this is the final design, you can consider conducting user testing or starting the final preparations before actual production.

[0070] As shown in FIG. 3, the interface diagram of the agent-based conceptual innovation design system provided in this embodiment is shown, and as shown in FIG. 4, the directory area is shown, which shows the project, design asset, design template, innovation method, and interactive creative tool.

[0071] The interface diagram provided in this embodiment also shows the canvas area, as shown in FIG. 5 and FIG. 6, which includes the analysis interface of the analysis process of the agent and the interaction interface between the planner and the designer, in which the innovation method can be added, the design asset can be added, the knowledge base can be added, and the template can be switched. The canvas area provided in this embodiment only shows the requirement analysis part of the analysis interface and the interaction interface between the planner and the designer, and the designer interacts with the large language model in the interaction interface, and the planner is the large language model.

[0072] The above description is only the preferred embodiment of the present application, it should be pointed out that for the ordinary skilled in the art, without departing from the principles of the present application, can also make several improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application. The structure, device and operation method not specifically described and explained in the present application, if no special description and limitation, are implemented according to the conventional means in the art.

Claims

1. A conceptual innovation design system based on intelligent agents, characterized in that, include: The conceptual innovation design module includes a graphical design process unit and a large language model dialogue unit. The graphical design process unit is used to visualize design assets and innovative methods to obtain a first design process. The large language model dialogue unit is used to interact with the user through a large language model. Based on the user's input information, the first design process is adjusted according to the process sequence of the design template to obtain a new design process. The new design process is stored in the design process creation unit. The final design asset, i.e., the design result, is obtained through the new design process. The design template management module is used to generate design templates and send them to the large language model. The innovation method module is used to store innovation methods and invoke innovation methods based on user input information to output new design assets. The design asset management module is used to store design assets and search for design assets that match the information entered by the user.

2. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The large language model dialogue unit is used to control the first design process according to the flow sequence of the design template based on the information input by the user to obtain a new design process. The control of the first design process includes adding new design assets, adding innovative methods, adding connections to existing design assets, adjusting the parameters of the function corresponding to the innovative method, and running one or more of the set innovative methods in the first design process. The information input by the user includes text descriptions, design assets, and innovative methods.

3. The agent-based conceptual innovation design system according to claim 1 or 2, characterized in that, The new design process is obtained by adjusting the first design process according to the process sequence of the design template, including: the large language model interacts with the user in real time through the process sequence of the design template, and adjusts the first design process in real time based on the information obtained during the interaction with the user.

4. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The design process graphical unit is used to graphically represent design assets and innovation methods to obtain a first design process. In the first design process, innovation methods and design assets are used as nodes. If the design asset is used as the input of the innovation method, a directed edge is set from the design asset to the innovation method. If the design asset is used as the output of the innovation method, a directed edge is set from the innovation method to the design asset. The design asset has a directed edge pointed to by the innovation method.

5. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The conceptual innovation design module also includes a design process management unit, which is used to create, open, edit names, delete and copy design processes.

6. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The conceptual innovation design module also includes a design asset traceability unit, which is used to save the design process corresponding to the generation of new design assets and convert the corresponding design process into a graphical representation for display to users.

7. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The innovation method module includes an innovation method display unit and an innovation method invocation unit; The innovation method display unit is used to search for innovation methods based on user purpose instructions. The innovation method includes innovation method type and innovation method parameters. The corresponding innovation method type is obtained based on the user purpose instructions. The innovation method parameters corresponding to the innovation method type are displayed to the user through a large language model. Based on the user's parameter selection instructions received through the large language model, the innovation method parameters are set to obtain an innovation method that matches the user purpose instructions. The innovation method invocation unit is used to obtain new design assets by running an adapted innovation method based on the design assets adapted to the user input information obtained from the design asset management module.

8. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The design asset management module includes a multimodal design asset storage unit, a cross-modal asset search unit, a design asset management unit, and a design asset import and export unit; The multimodal design asset storage unit is used to save design assets of different modalities in the form of binary data; The cross-modal asset search unit is used to convert multimodal design assets into vectors to obtain a vector database. During the search process, the user input information is converted into user vectors, and the user vectors are matched with the vectors in the vector database to obtain vectors with high matching degree and corresponding IDs. Based on the IDs, design assets that match the user input information are found. The design asset management unit is used to copy, delete, name edit, and view design assets; The design asset import and export unit is used to import design assets into external devices or download design assets to external devices.

9. The conceptual innovation design system based on intelligent agents according to claim 1, characterized in that, The design template management module includes a design template generation unit and a design template display unit; The design template generation unit is used to input the first design process into the large language model to obtain the corresponding text description, construct a design template through the first design process and the corresponding text description, and send the design template to the large language model. The design template display unit is used to graphically represent the first design process, represent the text description in text form, and display the graphical first design process and the text description in text form to the user through a large language model.

10. A method for implementing innovative conceptual design based on intelligent agents, characterized in that, The conceptual innovation design system based on intelligent agents according to any one of claims 1-9 obtains design results based on user input information, including: receiving user input information in real time using the large language model, adjusting the existing design process according to the process sequence of the design template based on the user input information to obtain a new design process, and completing the conceptual innovation design based on the final design asset, i.e. the design result, of the new design process.

Images