Meta-universe interior design system based on generative AI and dynamic optimization method

Abstract

The application discloses a meta-universe interior design system and a dynamic optimization method based on generative AI, aiming to solve the problem of low fusion degree of generative AI and meta-universe technology and lack of dynamic optimization capability of design scheme in the prior art. The system comprises user interaction, generative AI design, meta-universe scene construction and other modules, and the method realizes design closed loop through demand standardization, basic data preprocessing, initial scheme generation, meta-universe scene presentation, interactive feedback collection and iterative optimization. The application combines the advantages of generative AI automatic design and meta-universe immersive interaction, optimizes the scheme through quantitative indexes and algorithms, improves the design efficiency and user satisfaction, and can be flexibly adapted to various indoor design scenes such as family, business and public, and has high practicability and popularization value.

Description

Technical Field

[0001] This invention relates to the fields of metaverse technology, artificial intelligence design, and interior decoration technology, specifically a metaverse interior design system and dynamic optimization method based on generative AI. Background Technology

[0002] As people's quality of life improves, the demand for personalized and intelligent interior design is growing. Traditional interior design relies on the accumulated experience of designers, resulting in long design cycles, high communication costs, and low efficiency in iterating solutions, making it difficult to fully meet the diverse needs of users. Furthermore, users often cannot visually perceive the spatial effect before the design is finalized, leading to a common discrepancy between the final design and their expectations.

[0003] In recent years, the rise of metaverse technology has offered the possibility of immersive experiences for interior design, while generative AI technology has provided technical support for the automated generation of design solutions. While there are some attempts to combine AI with interior design and related applications of metaverse virtual scene display, most suffer from low integration and a lack of dynamic optimization capabilities for the design solutions. For example, some AI design systems can only generate static design solutions and cannot achieve immersive user interaction and feedback through metaverse scenes; while some metaverse interior display systems rely on manually preset scenes and cannot dynamically generate and optimize design solutions based on user needs.

[0004] Furthermore, current technologies rely heavily on manual judgment for design optimization, lacking quantifiable evaluation metrics and scientific optimization algorithms, resulting in low efficiency and unstable results. Therefore, how to deeply integrate the automatic design capabilities of generative AI with the immersive interactive capabilities of Metaverse to achieve dynamic optimization of interior design solutions, thereby improving design efficiency and user satisfaction, has become a pressing technical challenge in the field of interior design. Summary of the Invention

[0005] The technical problem this invention aims to solve is how to deeply integrate the automatic design capabilities of generative AI with the immersive interactive capabilities of metaverse to achieve dynamic optimization of interior design schemes, thereby improving design efficiency and user satisfaction.

[0006] The technical solution adopted in this invention is: a metaverse interior design system based on generative AI, comprising: a user interaction module, a generative AI design module, a metaverse scene construction module, a data acquisition and preprocessing module, and a dynamic optimization module;

[0007] The user interaction module is used to receive interior design requirement information input by the user, including space type, area parameters, style preference, functional requirements and budget threshold, and convert the design requirement information into standardized requirement data.

[0008] The data acquisition and preprocessing module is used to collect basic data related to interior design, including apartment layout data, building material attribute data, furniture model data, and lighting environment data. The module cleans, normalizes, and converts the basic data to generate a standardized basic dataset.

[0009] The generative AI design module, based on a preset generative AI model, takes into account the standardized requirements data and standardized basic dataset, and generates at least one initial interior design scheme. The initial interior design scheme includes three-dimensional spatial layout, furniture placement, building material selection, and lighting configuration.

[0010] The metaverse scene construction module is used to map the initial interior design scheme into a metaverse virtual interior scene, allowing users to enter the virtual interior scene through a virtual avatar for immersive browsing and interactive operations;

[0011] The dynamic optimization module is used to collect user interaction feedback data in the metaverse virtual indoor scene, combine it with preset optimization evaluation indicators, iterate and optimize the initial indoor design scheme, and output the final indoor design scheme and the corresponding metaverse virtual scene file.

[0012] As a further aspect of the present invention: the generative AI model is a fine-tuning model based on the Transformer architecture, and a loss function is used to control the training accuracy during the training process of the fine-tuning model. The loss function is: ;in, This represents the total loss value, used to measure the deviation between the model's predicted output and the actual design scheme, and to guide the adjustment of model parameters; , These are weighting coefficients, all ranging from (0,1), used to balance the effects of different loss terms; To ensure reasonable loss items are allocated, The loss term is style matching; the training dataset of the fine-tuning model includes historical interior design scheme data, metaverse scene adaptation data, and user preference tag data, and the output dimension of the generative AI model matches the coordinate dimension and attribute dimension of the three-dimensional interior design space.

[0013] As a further aspect of the present invention: the user interaction module also supports multimodal input methods, including text input, voice input, image reference input and gesture interaction input, and the multimodal input information is converted into unified standardized requirement data after semantic parsing.

[0014] As a further aspect of the present invention: the metaverse scene construction module includes a scene rendering unit, a physics engine unit, and a network synchronization unit; the scene rendering unit uses real-time ray tracing technology to achieve realistic rendering of the virtual indoor scene; the physics engine unit is used to simulate the mechanical stability of furniture placement and physical collision detection for spatial passage; the network synchronization unit supports multiple users to simultaneously access the metaverse virtual indoor scene for collaborative browsing and design discussions.

[0015] As a further aspect of the present invention: the optimization evaluation indicators of the dynamic optimization module include space utilization rate, functional adaptability, visual aesthetics, user satisfaction, and cost control rate. A weighted summation method is used to calculate the comprehensive evaluation score for each indicator. ;in, The score is used to comprehensively evaluate and determine whether the design scheme meets the optimization termination conditions; To determine the number of evaluation indicators, n=5 in this scheme; Let the weight of the i-th indicator satisfy the following condition: This is used to reflect the importance of different indicators; Let be the score for the i-th indicator, with a value range of [0, 100]. The dynamic optimization module iteratively updates the design scheme through incremental training of a generative AI model, and outputs the comprehensive evaluation score of the optimized scheme after each iteration. until ( The preset score threshold has a range of [80, 95].

[0016] As a further aspect of the present invention, it also includes a data storage module, which is used to store standardized requirement data, standardized basic dataset, initial interior design scheme, interactive feedback data, final interior design scheme, and metaverse virtual scene files. The data storage module adopts a distributed storage architecture, which supports fast query and backup of design data.

[0017] A metaverse interior design dynamic optimization method based on generative AI includes the following steps:

[0018] S1: Obtain the user's interior design requirements information through the user interaction module, standardize the design requirements information, and generate standardized requirements data;

[0019] S2: Collect basic interior design data through the data acquisition and preprocessing module, clean, normalize and convert the basic data to generate a standardized basic dataset;

[0020] S3: Input the standardized requirements data and standardized basic dataset into the generative AI model of the generative AI design module to generate at least one initial interior design scheme;

[0021] S4: The initial interior design scheme is mapped to a metaverse virtual interior scene through the metaverse scene construction module, providing users with an immersive browsing and interactive operation entry point;

[0022] S5: Collect user interaction feedback data in the metaverse virtual indoor scene, including browsing trajectory, dwell time, modification operations and evaluation rating;

[0023] S6: The dynamic optimization module iteratively optimizes the initial interior design scheme based on interactive feedback data and preset optimization evaluation indicators to generate an optimized design scheme;

[0024] S7: Determine whether the evaluation score of the optimized design scheme has reached the preset threshold. If not, return to step S6 to continue iterating; if it has, output the final interior design scheme and the corresponding metaverse virtual scene file.

[0025] As a further aspect of the present invention: when generating the initial interior design scheme in step S3, the generative AI model adopts a multi-objective generation strategy, simultaneously satisfying the rationality of spatial layout, style unity, and cost controllability. Cost controllability is constrained by the cost deviation rate. ;in, Cost deviation rate, used to measure the degree of deviation between the estimated cost of the design and the user's budget; The estimated total cost of the design scheme, The budget threshold input by the user is required. ( A preset deviation threshold is set, with a value range of [5%, 10%]). If the threshold is not met, a new scheme will be generated. After the initial interior design scheme is output, it must undergo compliance verification, which includes compliance with building design codes, fire safety standards, and indoor environmental protection standards.

[0026] As a further aspect of the present invention: the collection of interactive feedback data in step S5 adopts a combination of real-time collection and batch processing, caches the user's real-time operation data, summarizes and analyzes the cached data at preset time intervals, and extracts the user's core needs and preferences features as the basis for optimization design.

[0027] As a further aspect of the present invention: the iterative optimization process in step S6 includes four sub-steps: scheme deconstruction, parameter adjustment, scheme reconstruction, and evaluation and verification; the scheme deconstruction is used to break down the various components and related parameters of the initial design scheme; the parameter adjustment is based on user feedback characteristics to correct the corresponding parameters, and the parameter correction amount satisfies: ;in, This is the parameter correction amount, used to determine the magnitude of parameter adjustment; The adjustment coefficient has a value range of (0,1) and is used to avoid sudden changes in the scheme due to excessive parameter adjustment; The user feedback feature function outputs a value that is positively correlated with the intensity of the user feedback. The scheme reconstruction regenerates the optimized complete design scheme through a generative AI model. The evaluation verification is used to calculate the evaluation score of the optimized scheme and determine whether the iteration termination condition is met.

[0028] The beneficial effects of this invention are:

[0029] 1. Achieve deep integration of generative AI and metaverse technology, breaking through the barriers of separate application of existing technologies: Generative AI design module automatically generates initial design schemes that meet user needs, and combines them with metaverse scene construction module to achieve immersive presentation of the schemes, allowing users to intuitively perceive the spatial effects. At the same time, based on the interactive feedback in the metaverse scene, the schemes are dynamically optimized, forming a closed loop of "design-experience-optimization", which greatly improves the collaboration and integrity of the design process.

[0030] 2. Improve design efficiency and iteration speed: Relying on the automatic design capabilities of generative AI models, designers can quickly generate multiple initial solutions that meet the requirements without having to manually conceive solutions from scratch, thus shortening the design cycle. At the same time, the dynamic optimization module realizes automatic iteration of solutions based on quantitative evaluation indicators and algorithms, replacing the traditional manual optimization mode, significantly improving optimization efficiency and reducing labor costs.

[0031] 3. Enhance user interaction experience and demand matching: Support multimodal user interaction methods to lower the threshold for user input; the metaverse immersive scene allows users to browse and interact with the design scheme from all angles, resulting in more accurate feedback; the optimization process is based on real user interaction data to ensure that the final solution meets the user's personalized needs, reduce the deviation between the solution and expectations, and reduce the risk of rework.

[0032] 4. Enhance the scientific rigor and reliability of the design scheme: Control the training accuracy of the generative AI model through a preset loss function, quantify the merits of the scheme using a weighted summation comprehensive evaluation index, and constrain the parameter adjustment process through scientific algorithms, so that the generation and optimization of the design scheme have a quantitative basis, avoid the bias of human subjective judgment, and improve the rationality, compliance and stability of the scheme.

[0033] 5. Expand the application scenarios and collaborative capabilities of interior design: The network synchronization unit of the metaverse scene supports multi-user collaborative browsing and discussion, which can be applied to collaborative design needs in multiple scenarios such as home decoration and commercial space design; the distributed architecture of the data storage module ensures the security and reliability of design data, facilitates the traceability, reuse and subsequent modification of the solution, and enhances the practicality and promotion value of the technology. Attached Figure Description

[0034] Figure 1 This is a core architecture diagram of the metaverse interior design system and dynamic optimization method based on generative AI of this invention.

[0035] Figure 2 This is a flowchart of the dynamic optimization method of the metaverse interior design system and dynamic optimization method based on generative AI of the present invention. Detailed Implementation

[0036] The present invention will be further described in detail below with reference to specific embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0037] The core architecture of the metaverse interior design system and dynamic optimization method based on generative AI described in this invention is unified, including a user interaction module, a generative AI design module, a metaverse scene construction module, a data acquisition and preprocessing module, a dynamic optimization module, and a data storage module. These modules work together to achieve a complete process of "requirement input - solution generation - virtual experience - feedback optimization - solution output". The following examples, through different application scenarios, illustrate in detail the specific application methods of the system. Example 1: Interior Design Scenario for a Family Residence

[0038] This embodiment addresses a user's personalized design needs for a 120㎡ three-bedroom, two-living-room apartment. The specific application process is as follows:

[0039] 1. Requirements Input Stage: Users submit text requirements ("Modern minimalist style, master bedroom needs to reserve space for a walk-in closet, budget 250,000 yuan"), voice requirements ("Living room needs good lighting, sofa background wall needs to be fully equipped with storage"), and reference images (3 renderings of a modern minimalist style living room) through the multimodal input method of the user interaction module; The user interaction module performs semantic parsing on the multimodal information and converts it into standardized requirement data, including space type (three bedrooms and two living rooms), area parameters (120㎡), style tag (modern minimalist), functional requirements (master bedroom walk-in closet, living room full wall storage, enhanced lighting), budget threshold (250,000 yuan), and style reference feature vector.

[0040] 2. Data Acquisition and Preprocessing: The data acquisition and preprocessing module automatically collects the structural data of the apartment (location of load-bearing walls, door and window dimensions, floor height of 2.8m), building material data suitable for the modern minimalist style (light gray latex paint, natural wood-colored solid wood flooring, white sintered stone, etc.), furniture model data (simple fabric sofa, floating TV cabinet, sliding door wardrobe, etc.), and lighting environment data (south-facing floor-to-ceiling windows in the living room, with an average daily lighting duration of 6 hours). The collected data is cleaned (damaged furniture models are removed), normalized (the coordinate system is unified to metric), and converted (the apartment CAD file is converted into 3D model data) to generate a standardized basic dataset.

[0041] 3. Initial Scheme Generation: The generative AI design module calls a fine-tuned model based on the Transformer architecture, inputs standardized requirements data and a standardized basic dataset, and generates three initial interior design schemes; a loss function is used during the generation process. Control accuracy (of which) , (Focusing on the rationality of the layout), the three schemes differ in the design of the living room storage cabinet and the opening method of the master bedroom walk-in closet.

[0042] 4. Metaverse Scene Construction and Interaction: The metaverse scene construction module maps the three initial schemes to metaverse virtual indoor scenes respectively. It uses real-time ray tracing technology to render the lighting effect of the living room and uses a physics engine to simulate the opening and closing of the sliding door of the walk-in closet and the passage space. Users create virtual avatars to enter the scene and immerse themselves in browsing the three schemes. They mark the living room storage cabinet design and the sliding door design of the master bedroom walk-in closet in Scheme 2 as "satisfactory" and make adjustment requests for the position of the desk in the secondary bedroom in Scheme 2 ("the desk needs to be closer to the window"). The system collects the user's browsing trajectory in real time (the longest stay in the Scheme 2 scene is 15 minutes), marking information and modification requests, forming interactive feedback data.

[0043] 5. Dynamic Optimization and Solution Output: The dynamic optimization module calculates the initial score of Solution 2 based on interactive feedback data and using comprehensive evaluation indicators. (Not reaching the preset threshold) ); Regarding the need to "adjust the position of the desk in the secondary bedroom", a parameter correction formula was used. ( (User feedback was moderate) The desk coordinate parameters were adjusted, and an optimized solution was regenerated. A new metaverse virtual scene was constructed for user verification. Users had no new requirements, and the overall score of the optimized solution was [not specified]. (Meets the threshold); Output the final design scheme (including 3D layout diagram, building material list, budget details of 248,000 yuan) and the metaverse virtual scene file (allowing users to view it at any time). Example 2: Interior Design Scenario for Commercial Office Space

[0044] This example addresses the design requirements of a 500㎡ open-plan office space for a technology company. The core requirements are "efficient collaboration, flexible zoning, a tech-inspired style, and support for 20 people working simultaneously." The specific application process is as follows:

[0045] 1. Requirements Input Phase: Enterprise administrative staff submit text requirements and departmental function lists through the user interaction module, and simultaneously invite 3 department heads to collaboratively input requirements through the network synchronization unit (the R&D department needs an independent discussion area, and the marketing department needs a display area); the user interaction module summarizes the requirements from multiple users and converts them into standardized requirement data, including space type (open office space), area parameters (500㎡), style tags (technological feel), functional requirements (20 open workstations, 1 R&D discussion area, 1 marketing display area, and a tea room), and budget threshold (800,000 yuan).

[0046] 2. Data Acquisition and Preprocessing: The data acquisition and preprocessing module collects apartment structure data (large span, no load-bearing walls, 3.5m floor height), technological style building material data (metallic ceiling, glass partition, gray industrial style floor tiles, etc.), office furniture model data (modular workstations, movable conference tables, electronic display screens, etc.), and lighting environment data (30 lighting points evenly distributed on the top). After preprocessing, a standardized basic dataset is generated, with key annotations on the power supply interface location data of electronic devices.

[0047] 3. Initial Solution Generation: The generative AI design module calls the fine-tuning model, inputs standardized requirements data and a basic dataset, and generates two initial solutions. The difference lies in the arrangement of the open workstations (row / column / matrix) and the location layout of the display area; the loss function is set to... Balancing layout rationality with style matching; and simultaneously controlling cost deviation rate To control costs, the estimated costs for the two plans are 780,000 yuan and 795,000 yuan respectively. All less than 8% (preset) This meets cost requirements.

[0048] 4. Metaverse Scene Construction and Interaction: The metaverse scene construction module builds virtual scenes for two different scenarios. The network synchronization unit supports simultaneous access and collaborative browsing for three department heads. The physics engine simulates the walking distance between workstations (ensuring ≥1.2m), and the scene rendering unit displays technological lighting effects (gradient LED light strips). During the interaction, the R&D department head suggested adding sound insulation to the discussion area, and the marketing department head approved the matrix workstation arrangement. The system collects multi-user interaction feedback data.

[0049] 5. Dynamic Optimization and Solution Output: The dynamic optimization module adjusts the solution based on multi-user feedback, adds sound insulation parameters for the glass partition in the discussion area, and recalculates the overall score. (Meets the threshold); Output the final solution (including workstation layout diagram, building material list, electronic equipment configuration table) and metaverse virtual scene file, which can be previewed by the enterprise's subsequent organization of employees. Example 3: Interior Design Scenario of a Public Exhibition Hall

[0050] This example addresses the design requirements of a 300㎡ themed exhibition hall in a city planning exhibition area. The core requirements are "showcasing the city's development history, providing an interactive experience, adopting a dignified and minimalist style, and accommodating large visitor volumes." The specific application process is as follows:

[0051] 1. Demand Input Stage: The exhibition hall operator submits textual requirements and a list of display content through the user interaction module, which are then converted into standardized demand data, including space type (themed exhibition hall), area parameters (300㎡), style tags (solemn and simple), functional requirements (historical timeline display area, interactive projection area, reception desk, emergency passage reservation), and budget threshold (1.2 million yuan). The "efficiency of pedestrian flow" is highlighted as the core evaluation indicator.

[0052] 2. Data Collection and Preprocessing: Collect apartment layout data (emergency passage on the left side of the entrance, large open space in the middle, and small areas that can be divided on the right side), data on dignified and minimalist style building materials (white marble walls, light gray stone flooring, etc.), data on display equipment models (interactive projection equipment, touch screen query screen, timeline display stand, etc.), and basic data on pedestrian flow simulation (estimated daily visitor count of 500 people); after preprocessing, generate a standardized basic dataset, which mainly includes the compliant size data of the emergency passage (≥1.4m).

[0053] 3. Initial Design Generation: The generative AI design module generates two initial designs, differing in the orientation of the timeline display area (straight line / curved) and the number of interactive projection areas; the loss function is set to... The focus is on the rationality of the layout (ensuring the evacuation of people); the cost deviation rate is controlled within 5% to meet the budget requirements.

[0054] 4. Metaverse Scene Construction and Interaction: The metaverse scene construction module builds virtual scenes to simulate the flow density of 500 people visiting at the same time (by setting the movement rules of the personnel models through the physics engine); the operators experience the visit through virtual avatars and provide feedback such as "the arc-shaped timeline display area is more in line with the visitor flow" and "one more interactive projection area is needed". The system collects the simulated flow data and user modification requests.

[0055] 5. Dynamic Optimization and Solution Output: The dynamic optimization module adjusts the solution to an arc-shaped timeline display area, adds an interactive projection area, and re-simulates pedestrian evacuation efficiency, with a comprehensive score. Output the final solution and virtual scene files, which can be used for subsequent construction briefings and promotion. Comparative Analysis of Examples

[0056] The three embodiments described above correspond to three typical interior design scenarios: residential, commercial, and public. Each embodiment differs in system module application, parameter settings, and emphasis on core requirements, as shown in the table below:

[0057]

[0058] The above comparison demonstrates that the system described in this invention can flexibly adapt to different types of interior design scenarios. By adjusting the parameters of the generative AI model (loss function weights, cost deviation thresholds, etc.) and the functional configuration of the metaverse scenario, it can meet the core needs of different scenarios. Specifically, the home scenario emphasizes personalized experience and cost control, the commercial office scenario emphasizes multi-user collaboration and office efficiency, and the public exhibition hall scenario emphasizes the rationality and compliance of traffic flow. Each embodiment achieves accurate generation of design solutions through a closed-loop process of "design-experience-optimization," verifying the versatility and practicality of the system of this invention.

[0059] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A metaverse interior design system based on generative AI, characterized in that, include: The module includes a user interaction module, a generative AI design module, a metaverse scene construction module, a data acquisition and preprocessing module, and a dynamic optimization module. The user interaction module is used to receive interior design requirement information input by the user, including space type, area parameters, style preference, functional requirements and budget threshold, and convert the design requirement information into standardized requirement data. The data acquisition and preprocessing module is used to collect basic data related to interior design, including apartment structure data, building material attribute data, furniture model data, and lighting environment data. The basic data is cleaned, normalized, and format converted to generate a standardized basic dataset. The generative AI design module, based on a preset generative AI model, takes into account the standardized requirements data and standardized basic dataset, and generates at least one initial interior design scheme. The initial interior design scheme includes three-dimensional spatial layout, furniture placement, building material selection, and lighting configuration. The metaverse scene construction module is used to map the initial interior design scheme into a metaverse virtual interior scene, allowing users to enter the virtual interior scene through a virtual avatar for immersive browsing and interactive operations; The dynamic optimization module is used to collect user interaction feedback data in the metaverse virtual indoor scene, combine it with preset optimization evaluation indicators, iterate and optimize the initial indoor design scheme, and output the final indoor design scheme and the corresponding metaverse virtual scene file.

2. The metaverse interior design system based on generative AI according to claim 1, characterized in that, The generative AI model is a fine-tuning model based on the Transformer architecture. During the training process of this fine-tuning model, a loss function is used to control the training accuracy. The loss function is: ;in, This represents the total loss value, used to measure the deviation between the model's predicted output and the actual design scheme, and to guide the adjustment of model parameters; , These are weighting coefficients, all ranging from (0,1), used to balance the impact of different loss terms; To ensure reasonable loss items are allocated, The loss term is style matching; the training dataset of the fine-tuning model includes historical interior design scheme data, metaverse scene adaptation data, and user preference tag data, and the output dimension of the generative AI model matches the coordinate dimension and attribute dimension of the three-dimensional interior design space.

3. The metaverse interior design system based on generative AI according to claim 1, characterized in that, The user interaction module also supports multimodal input methods, including text input, voice input, image reference input, and gesture interaction input. The multimodal input information is converted into unified standardized requirement data after semantic parsing.

4. The metaverse interior design system based on generative AI according to claim 1, characterized in that, The metaverse scene construction module includes a scene rendering unit, a physics engine unit, and a network synchronization unit. The scene rendering unit uses real-time ray tracing technology to achieve realistic rendering of virtual indoor scenes. The physics engine unit is used to simulate the mechanical stability of furniture placement and physical collision detection for spatial passage. The network synchronization unit supports multiple users to simultaneously access the metaverse virtual indoor scene for collaborative browsing and design discussions.

5. The metaverse interior design system based on generative AI according to claim 1, characterized in that, The optimization evaluation indicators of the dynamic optimization module include space utilization, functional adaptability, visual aesthetics, user satisfaction, and cost control rate. The comprehensive evaluation score is calculated by weighted summation of each indicator. ;in, The score is used to comprehensively evaluate and determine whether the design scheme meets the optimization termination conditions; To determine the number of evaluation indicators, n=5 in this scheme; Let the weight of the i-th indicator satisfy the following condition: This is used to reflect the importance of different indicators; Let be the score for the i-th indicator, with a value range of [0, 100]. The dynamic optimization module iteratively updates the design scheme through incremental training of a generative AI model, and outputs the comprehensive evaluation score of the optimized scheme after each iteration. until ( The preset score threshold has a range of [80, 95].

6. The metaverse interior design system based on generative AI according to claim 1, characterized in that, It also includes a data storage module, which is used to store standardized requirement data, standardized basic datasets, initial interior design schemes, interactive feedback data, final interior design schemes, and metaverse virtual scene files. The data storage module adopts a distributed storage architecture, which supports fast querying and backup of design data.

7. A metaverse interior design dynamic optimization method based on generative AI, characterized in that, The method applied to the metaverse interior design system based on generative AI as described in any one of claims 1-6 includes the following steps: S1: Obtain the user's interior design requirements information through the user interaction module, standardize the design requirements information, and generate standardized requirements data; S2: Collect basic interior design data through the data acquisition and preprocessing module, clean, normalize and convert the basic data to generate a standardized basic dataset; S3: Input the standardized requirements data and standardized basic dataset into the generative AI model of the generative AI design module to generate at least one initial interior design scheme; S4: The initial interior design scheme is mapped to a metaverse virtual interior scene through the metaverse scene construction module, providing users with an immersive browsing and interactive operation entry point; S5: Collect user interaction feedback data in the metaverse virtual indoor scene, including browsing trajectory, dwell time, modification operations and evaluation rating; S6: The dynamic optimization module iteratively optimizes the initial interior design scheme based on interactive feedback data and preset optimization evaluation indicators to generate an optimized design scheme; S7: Determine whether the evaluation score of the optimized design scheme has reached the preset threshold. If not, return to step S6 to continue iterating; if it has, output the final interior design scheme and the corresponding metaverse virtual scene file.

8. The metaverse interior design dynamic optimization method based on generative AI according to claim 7, characterized in that, In step S3, when generating the initial interior design scheme, the generative AI model adopts a multi-objective generation strategy, simultaneously satisfying the requirements of spatial layout rationality, style consistency, and cost controllability. Cost controllability is constrained by the cost deviation rate. ;in, Cost deviation rate, used to measure the degree of deviation between the estimated cost of the design and the user's budget; The estimated total cost of the design scheme, The budget threshold input by the user is required. ( A deviation threshold is set, with a value range of [5%, 10%]). If the threshold is not met, a new scheme is generated. After the initial interior design scheme is output, it must undergo compliance verification, which includes compliance with building design codes, fire safety standards, and indoor environmental protection standards.

9. The metaverse interior design dynamic optimization method based on generative AI according to claim 7, characterized in that, In step S5, the collection of interactive feedback data adopts a combination of real-time collection and batch processing. The user's real-time operation data is cached, and the cached data is summarized and analyzed at preset time intervals to extract the user's core needs and preferences, which serves as the basis for optimization design.

10. The metaverse interior design dynamic optimization method based on generative AI according to claim 7, characterized in that, Step S6, the iterative optimization process, includes four sub-steps: scheme deconstruction, parameter adjustment, scheme reconstruction, and evaluation and verification. Scheme deconstruction involves breaking down the initial design scheme into its constituent parts and related parameters. Parameter adjustment corrects the corresponding parameters based on user feedback characteristics, with the correction amount satisfying the following: ;in, This is the parameter correction amount, used to determine the magnitude of parameter adjustment; The adjustment coefficient has a value range of (0,1) and is used to avoid sudden changes in the scheme due to excessive parameter adjustment; The user feedback feature function outputs a value that is positively correlated with the intensity of the user feedback. The scheme reconstruction regenerates the optimized complete design scheme through a generative AI model. The evaluation verification is used to calculate the evaluation score of the optimized scheme and determine whether the iteration termination condition is met.

Images