CN-121834990-B - Multi-mode data-driven AI house type customization scheme generation method and system

CN121834990BCN 121834990 BCN121834990 BCN 121834990BCN-121834990-B

Abstract

The application relates to a multi-mode data-driven AI house type customization scheme generation method and system, comprising the steps of acquiring multi-mode data and extracting features to obtain a design target feature set, matching through a building standard knowledge base to generate a preliminary constraint condition set, matching a plurality of optimization strategies, generating a reinforced constraint condition set through a first agent, generating a space planning scheme set through a second agent, generating a stylized layout scheme set through a third agent, generating a parameterized building information model set through a fourth agent, and obtaining multi-mode data and extracting features, matching and constraining through the building standard knowledge base, combining agent optimization to generate a reinforced constraint condition, further generating a space planning scheme, a stylized scheme and a parameterized building information model, and solving the problem that the prior art lacks flexibility, controllability and engineering compliance through multi-mode data driving and agent optimization.

Inventors

ZHAO JIE
ZHAO PENG
Liu Chouzhong
HUANG YUANZHONG
ZHANG LIMING
LI JUN
YANG FENGGUO
YAN YIFENG

Assignees

中京科技(广州)有限公司

Dates

Publication Date: 20260508
Application Date: 20260313

Claims (8)

1. A multi-mode data-driven AI house type customization scheme generation method is characterized by comprising the following steps: responding to the received task instruction, acquiring corresponding multi-mode data, and carrying out feature extraction on the task instruction and the multi-mode data to obtain a design target feature set, wherein the design target feature set comprises a user preference vector, an optimization target identifier and a constraint condition identifier; matching constraint condition identifiers through a pre-constructed building specification knowledge base to generate a preliminary constraint condition set; matching a plurality of optimization strategies from a preset optimization strategy pool based on the optimization target identifier, and carrying out multi-strategy collaborative optimization on the preliminary constraint condition set through the combination of the matched optimization strategies and the user preference vector by the first agent to generate a reinforced constraint condition set; inputting the reinforced constraint condition set into a second agent to generate a space planning scheme set, and generating a stylized layout scheme set by combining the space planning scheme set and a user preference vector through a third agent; inputting a stylized layout scheme set to a fourth agent to generate a parameterized building information model set, wherein the fourth agent specifically comprises the steps of identifying building component characteristics and indoor element characteristics in the stylized layout scheme set, and extracting attribute information, wherein the attribute information comprises category labels, geometric parameters, spatial positions and associated attributes; The fourth agent matches corresponding parameterized component families from a predefined parameterized component library based on building component features and indoor element features, and generates a plurality of three-dimensional parameterized components in an instantiation mode according to the extracted attribute information; the fourth agent generates a component connection relationship based on the three-dimensional parameterized component and the assembly logic thereof, wherein the component connection relationship comprises a space topological relationship, a hierarchical assembly relationship and an engineering constraint relationship; The fourth agent generates a parameterized building information model set based on the three-dimensional parameterized components and the component connection relations; The second agent includes a constraint analysis layer, a space optimization layer and a scheme decoding layer, the reinforced constraint condition set is input to the second agent, and a space planning scheme set is generated, including: The constraint analysis layer analyzes the reinforced constraint condition set, extracts space function constraint features, area constraint features and topological relation constraint features, and generates space parameterized representation; the space optimization layer executes a preset iterative optimization flow; The scheme decoding layer generates a space planning scheme set comprising at least one main scheme and a plurality of variant schemes based on the space parameterized representation obtained after the execution of the iterative optimization flow and the intermediate optimization state recorded in the iterative process; the third agent includes style coding layer, layout condition generating layer, physical checking layer and adjusting processing layer, the generating stylized layout scheme set by combining space planning scheme set and user preference vector by the third agent includes: the style coding layer maps the user preference vector into a style potential vector; The layout condition generation layer encodes the space planning scheme set into corresponding space structure condition vectors, and fuses the style potential vectors and the space structure condition vectors to generate a preliminary layout scheme; the physical verification layer performs collision detection and compliance verification on the preliminary layout scheme based on a predefined physical rule base and a furniture knowledge base to generate a verification report; the adjustment processing layer adjusts the preliminary layout scheme based on the verification report and generates a stylized layout scheme set.
2. The method for generating a multi-modal data-driven AI house type customization scheme as set forth in claim 1, wherein said steps of responding to the received task instruction, obtaining corresponding multi-modal data, and performing feature extraction on the task instruction and the multi-modal data to obtain a design target feature set, said design target feature set including a user preference vector, an optimization target identifier, and a constraint condition identifier, include the steps of: inputting the task instruction and the multi-mode data into a pre-trained feature extraction model to obtain a fusion semantic feature; carrying out parallel analysis on the fusion semantic features to generate a preliminary preference vector, an initial target identifier and an initial constraint identifier; and carrying out consistency check and calibration on the preliminary preference vector, the initial target identifier and the initial constraint identifier to obtain a design target feature set.
3. The method for generating the multi-modal data-driven AI house type customization scheme as set forth in claim 2, wherein the multi-modal data includes a first data and a second data, the feature extraction model includes a feature encoding layer, a feature fusion layer and a feature enhancement layer, and the step of inputting the task instruction and the multi-modal data into the pre-trained feature extraction model to obtain the fused semantic features includes the steps of: The feature encoding layer encodes first data to generate text feature vectors, space feature vectors and parameter feature vectors, wherein the first data comprises task text information, house type structure information and structure parameter information; the feature fusion layer splices and fuses the text feature vector, the space feature vector and the parameter feature vector to generate a fusion feature vector; The feature enhancement layer encodes the second data to generate an enhancement feature vector, and performs feature recalibration on the fusion feature vector by taking the enhancement feature vector as a modulation signal to generate fusion semantic features.
4. The method for generating the multi-modal data-driven AI house type customization scheme as set forth in claim 2, wherein said step of generating the preliminary preference vector, the initial target identifier, and the initial constraint identifier by parsing the fusion semantic features in parallel includes the steps of: inputting the fusion semantic features into a pre-constructed parallel analysis model, wherein the parallel analysis model comprises a first analysis layer, a second analysis layer and a third analysis layer; the first analysis layer maps the fusion semantic features to a continuous preference semantic space to generate a preliminary preference vector; The second analysis layer carries out multi-label classification on the fusion semantic features, and identifies at least one dominant optimization target from a predefined optimization target set to generate a corresponding initial target identifier; And the third analysis layer performs sequence labeling and key information extraction on the fusion semantic features, identifies physical limiting conditions and functional limiting conditions, maps the physical limiting conditions and the functional limiting conditions into standard constraint entries in a predefined constraint library, and generates an initial constraint identifier.
5. The method for generating a multi-modal data-driven AI house type customization scheme as set forth in claim 1 wherein said step of matching constraint identifiers through a pre-constructed building specification knowledge base to generate a preliminary set of constraints comprises the steps of: Analyzing constraint condition identifiers, identifying constraint types and constraint parameters, reasoning based on a preset association rule set, adding implicit constraint conditions based on a reasoning result, and generating a constraint identifier list containing a plurality of constraint identifiers; Traversing the constraint identifier list, inquiring parameterization rules, value ranges and boundary conditions corresponding to all constraint identifiers in a building specification knowledge base, and instantiating each constraint identifier into at least one atomic constraint condition based on an inquiry result; and carrying out logic conflict detection on all the atom constraint conditions, and if the conflict is detected, carrying out adjustment and optimization on the atom constraint conditions with the conflict according to constraint priority rules and conflict resolution strategies preset in a building specification knowledge base to generate a preliminary constraint condition set.
6. The method for generating the multi-modal data-driven AI house type customization scheme as set forth in claim 1, wherein the step of matching a plurality of optimization strategies from a preset optimization strategy pool based on the optimization target identifier, and performing multi-strategy collaborative optimization on the preliminary constraint condition set through the combination of the matched optimization strategies and the user preference vector by the first agent, and generating the reinforcement constraint condition set comprises the steps of: Dividing the preliminary constraint condition set to obtain a rigid constraint set and a flexible constraint set; inputting the matched optimization strategy, the user preference vector and the preliminary constraint condition set into a first agent, and enabling the first agent to execute a hierarchical optimization flow, wherein the hierarchical optimization flow comprises: Maintaining the rigid constraint set unchanged, constructing a multi-objective optimization problem with the flexible constraint set as an optimization variable, carrying out iterative solution on the multi-objective optimization problem, and calculating the pareto optimal relaxation interval of the flexible constraint set on the premise of meeting the rigid constraint set, wherein the optimization objective of the multi-objective optimization problem is defined by a matched optimization strategy; Based on the user preference vector, selecting a corresponding re-constraint value for each flexible constraint in the flexible constraint set in the pareto optimal relaxation interval, and generating a constraint adjustment scheme; and updating parameters of the flexible constraint set according to the constraint adjustment scheme to generate an intensified constraint condition set.
7. The method for generating the multi-modal data-driven AI house type customization scheme as set forth in claim 1, wherein the iterative optimization process comprises: (a) Calculating the matching loss of the space parameterized representation and the strengthening constraint condition set; (b) Based on the matching loss, updating the space parameterized representation through a preset back propagation algorithm; (c) And (3) taking the steps (a) and (b) as an iteration process, repeating the iteration process until any termination condition is reached, and recording an intermediate optimization state generated in the iteration process, wherein the termination condition comprises that the matching loss converges to a preset loss threshold value and the repetition frequency of the iteration process reaches the maximum iteration frequency.
8. A multi-modal data-driven AI home type customization scheme generating system for performing the steps of a multi-modal data-driven AI home type customization scheme generating method as claimed in any one of claims 1 to 7, comprising: the feature extraction module is used for responding to the received task instruction, obtaining corresponding multi-mode data, and extracting features of the task instruction and the multi-mode data to obtain a design target feature set, wherein the design target feature set comprises a user preference vector, an optimization target identifier and a constraint condition identifier; the constraint matching module is used for matching constraint condition identifiers through a pre-constructed building specification knowledge base to generate a preliminary constraint condition set; the constraint enhancement module is used for matching a plurality of optimization strategies from a preset optimization strategy pool based on the optimization target identifier, and carrying out multi-strategy collaborative optimization on the preliminary constraint condition set through the combination of the matched optimization strategies and the user preference vector by the first agent to generate an enhanced constraint condition set; the scheme generation module is used for inputting the reinforced constraint condition set into the second agent to generate a space planning scheme set, and generating a stylized layout scheme set by combining the space planning scheme set and the user preference vector through the third agent; And the model generation module is used for inputting the stylized layout scheme set to the fourth agent and generating a parameterized building information model set.

Description

Multi-mode data-driven AI house type customization scheme generation method and system Technical Field The application relates to the technical field of decoration design, in particular to a multi-mode data-driven AI house type customization scheme generation method and system. Background In the fields of indoor design, residence planning and decoration, efficient and accurate generation of customized house type schemes meeting the diversified and personalized requirements of users is always a core challenge in the industry field. With the evolution of computer aided design technology, an automatic intelligent house type generation system is developed, but has a certain defect in practical application. At present, the prior art mainly shows two paths, namely a system based on a template library and rule driving, wherein a large number of standardized house type templates are generally embedded in the system, so that a user can only combine limited operations such as size adjustment or component replacement, the operation logic of the system depends on preset hard rules, such as space area proportion and basic line specification, the scheme library scale directly restricts the upper limit of output diversity, when the user demand deviates from the preset templates or rules, the system often lacks self-adaptive capacity, complicated secondary adjustment is needed to be carried out depending on manual intervention, the urgent demands of the market on uniqueness and personalized design are difficult to respond, and secondly, a method based on a generated model is used for directly generating an indoor layout image with novel vision by learning massive house type map data, however, the method is difficult to effectively embed building specifications, structural safety requirements and accurate user function constraints due to the fact that the generated result has a certain innovation, but engineering rigidity requirements such as space utilization rate and bearing structure are easily ignored, and the scheme is only stopped at a construction reference stage, and practical and ground application standards cannot be met. Disclosure of Invention In order to solve the defects, the application provides a multi-mode data-driven AI house type customization scheme generation method and system. The first object of the present application is achieved by the following technical solutions: a multi-mode data-driven AI house type customizing scheme generating method comprises the following steps: responding to the received task instruction, acquiring corresponding multi-mode data, and carrying out feature extraction on the task instruction and the multi-mode data to obtain a design target feature set, wherein the design target feature set comprises a user preference vector, an optimization target identifier and a constraint condition identifier; matching constraint condition identifiers through a pre-constructed building specification knowledge base to generate a preliminary constraint condition set; matching a plurality of optimization strategies from a preset optimization strategy pool based on the optimization target identifier, and carrying out multi-strategy collaborative optimization on the preliminary constraint condition set through the combination of the matched optimization strategies and the user preference vector by the first agent to generate a reinforced constraint condition set; inputting the reinforced constraint condition set into a second agent to generate a space planning scheme set, and generating a stylized layout scheme set by combining the space planning scheme set and a user preference vector through a third agent; and inputting the stylized layout scheme set into a fourth agent to generate a parameterized building information model set. The second object of the present application is achieved by the following technical solutions: a multi-modal data-driven AI house type customization scheme generation system, comprising: the feature extraction module is used for responding to the received task instruction, obtaining corresponding multi-mode data, and extracting features of the task instruction and the multi-mode data to obtain a design target feature set, wherein the design target feature set comprises a user preference vector, an optimization target identifier and a constraint condition identifier; the constraint matching module is used for matching constraint condition identifiers through a pre-constructed building specification knowledge base to generate a preliminary constraint condition set; the constraint enhancement module is used for matching a plurality of optimization strategies from a preset optimization strategy pool based on the optimization target identifier, and carrying out multi-strategy collaborative optimization on the preliminary constraint condition set through the combination of the matched optimization strategies and the user preference vector by the first agent to generate an e