A multi-modal digital intelligent collaborative management and control and intelligent expression system for urban design

Abstract

The application discloses a kind of multi-modal digitalization collaborative management and intelligent expression system for urban design, it is related to wisdom city construction field, including: analysis module, for receiving city design related natural language interaction instruction, extraction design demand, constraint condition and target parameter, to the three-dimensional data of extraction is structured translation, to output structured design parameter;The present application translates city design natural language instruction into structured design parameter and matches city space basic geographic data to build accurate base model, so that design demand is efficiently landed, and the scheme is checked in real time with all elements in compliance with the combination of city design specification knowledge base, to ensure that design scheme fits legal control requirements, based on compliant scheme to complete block refinement, skin generation and scene layout and automatically generate effect picture and roaming animation, improve the visual expression efficiency of city design achievement.

Description

Technical Field

[0001] This invention relates to the field of smart city construction technology, specifically to a multimodal digital intelligent collaborative management and intelligent expression system for urban design. Background Technology

[0002] Urban design is a core component supporting the implementation of green buildings and promoting the high-quality construction of smart cities. Currently, the industry is accelerating its transformation towards digitalization and intelligence. Technologies such as multimodal information understanding, retrieval-enhanced generation, parametric modeling, and AIGC rendering have gradually permeated various design scenarios. As the demand for more refined urban planning increases, the industry's requirements for the collaboration and intelligence level of the entire design process are rising. Various basic intelligent tools have been applied in single stages such as modeling and drawing, driving the design mode from traditional manual dominance to semi-intelligentization.

[0003] The invention patent application with application number 202210969295.8 discloses a method for digitally generating building volumes based on morphological types in urban design. This application aims to solve the problem that "at present, the three-dimensional shape effect display in most urban designs in China relies on manual modeling, which requires a lot of time and effort to complete. Moreover, it mostly follows the process from the basic plan to the three-dimensional volume modeling. Therefore, the display of the three-dimensional shape effect is often lagging behind. When problems are found and the scheme needs to be adjusted, it is still necessary to go back to the two-dimensional plan design and then to the three-dimensional modeling process, and multiple modifications and optimizations are time-consuming and labor-intensive."

[0004] However, existing technologies still suffer from several problems: they cannot connect the entire urban design process; data silos are prominent; multimodal and parametric modeling and AIGC rendering functions are fragmented; there is neither natural language interaction to drive design iteration nor standardized real-time verification; all of which seriously restrict design efficiency and planning quality.

[0005] To address this, we propose a multimodal digital collaborative management and intelligent expression system for urban design. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a multimodal digital intelligent collaborative management and intelligent expression system for urban design, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions; This invention discloses a multimodal digital intelligent collaborative management and intelligent expression system for urban design, comprising: The system comprises four modules: a parsing module (receiving natural language interaction commands related to urban design, extracting design requirements, constraints, and target parameters, and translating the extracted 3D data into structured parameters); a construction module (calling urban spatial basic geographic data based on the structured design parameters to build an urban design base model and match spatial coordinates for the target site); a generation module (calling a preset design case knowledge base based on the completed base model and structured design parameters to generate design concepts, extrapolate architectural forms, and iteratively optimize schemes); a verification module (connecting to a preset urban design standard knowledge base to perform real-time verification of all constraints on the iteratively generated design schemes, outputting the verification results and synchronizing them to the generation module); an expression module (receiving verified design schemes, simultaneously refining the blocks, generating the skin, and setting up background elements, and driving the system's built-in AIGC engine to automatically generate renderings and walkthrough animations); and a management module (linking and storing the entire process data of the parsing, construction, generation, and verification modules with the output results of the expression module, setting the output results of the expression module as a unique data retrieval index). The parsing module is interconnected with the construction module via a wireless network. The construction module is interconnected with the generation module via a wireless network. The generation module is interconnected with the verification module via a wireless network. The verification module is interconnected with the expression module and the management module via a wireless network.

[0008] Furthermore, after receiving natural language interaction instructions related to urban design, the parsing module performs word segmentation and semantic entity recognition on the instruction text, splitting it into design requirement entities corresponding to design requirements, constraint entities corresponding to constraints, and target parameter entities corresponding to target parameters. Then, it performs semantic disambiguation and hierarchical mapping on each type of entity, mapping the unstructured natural language semantics to the preset urban design full-element structured parameter spectrum. It performs structured translation on the extracted three-dimensional data and outputs structured design parameters that correspond one-to-one with the preset urban design full-element structured parameter spectrum.

[0009] Furthermore, the structured design parameters include at least the target plot spatial boundary parameters, spatial coordinate system parameters, land use parameters, design control index parameters, and design target parameters; The construction module, based on the target plot spatial boundary parameters in the structured design parameters, calls urban spatial basic geographic data to extract topographic elevation data, land ownership data, existing building and structure data and municipal supporting data of the target plot and its surrounding preset range, and completes the layered construction of the base model; at the same time, based on the spatial coordinate system parameters in the structured design parameters, the completed base model is subjected to coordinate system origin calibration and spatial scale normalization processing to match the base model with the overall urban spatial coordinate system.

[0010] Furthermore, based on the completed base model and structured design parameters, the generation module first uses the land use, spatial scale, and design objectives of the target plot as core matching dimensions. It then matches design cases from a pre-defined design case knowledge base whose comprehensive similarity across all dimensions meets a preset threshold. The module extracts the core elements of the design concept and architectural form control rules from these cases. Next, it performs an adaptation mapping between the extracted elements and rules and the structured design parameters to generate the initial design scheme. Finally, using spatial adaptability as the core optimization objective, it performs architectural form deduction and iterative scheme optimization. The spatial adaptability... ; In the formula: This is the ratio of the actual floor area ratio of the proposed scheme to the target floor area ratio given in the structured design parameters. It is the minimum ratio of the actual setback distance of the building foundation in each direction to the minimum setback distance of the corresponding direction constraint; K is the ratio of the absolute value of the difference between the highest building height of the proposed scheme and the average height of existing buildings within the surrounding preset range to the preset height threshold; K is the ratio of the actual building density of the proposed scheme to the target building density given in the structural design parameters. During the iterative optimization process, when the spatial fit S reaches the preset threshold, the iteration stops and the optimized design solution is output.

[0011] Furthermore, after the verification module connects to a preset urban design code knowledge base, it first divides the code clauses in the knowledge base into two categories: mandatory code clauses and recommended code clauses. Then, for each iteratively generated design scheme, it performs a compliance determination on each of the two types of clauses, and outputs the code compliance score of the scheme. ; In the formula: This refers to the total number of mandatory regulatory clauses in the pre-defined urban design code knowledge base that match the target plot. This refers to the total number of recommended code clauses in the pre-defined urban design code knowledge base that match the target plot. For the first The compliance coefficient of each mandatory regulatory clause; For the first The compliance coefficient of the recommended normative clause is 1 when the solution meets the requirements of the clause, and 1 when it does not meet the requirements. The verification module will standardize compliance. The calculation results are synchronized to the generation module as verification results. The generation module adjusts the direction and step size of the scheme iteration optimization based on the verification results.

[0012] Furthermore, after receiving the verification results, the generation module determines the compliance status of the solution based on the calculation results of the compliance degree. When the compliance coefficient of any mandatory standard clause is 0, the generation module locks the design element constrained by the corresponding non-compliant mandatory standard clause. Taking the elimination of the non-compliance of the mandatory standard clause as the only direction of iterative optimization, the iteration step size is adjusted to the preset minimum step size. Simultaneously, the boundary parameter threshold corresponding to the non-compliant mandatory standard clause is extracted to determine the compliance value range of the locked design element. The parameter value of the design element in the current solution is used as the adjustment starting point, and the preset minimum step size is used as the single adjustment unit. The adjustment is carried out unidirectionally and successively within the compliance value range. After each adjustment, only the compliance verification is re-performed for the mandatory standard clause. During the adjustment process, other non-locked compliant design elements are not modified until the compliance coefficient of the mandatory standard clause changes to 1, so as to complete the correction of all non-compliant mandatory standard clauses. When the compliance coefficient of all mandatory regulatory clauses is 1, the generation module takes improving the compliance to a preset threshold as its core objective. It obtains all unmet recommended regulatory clauses, sorts them by their corresponding compliance discount coefficients from smallest to largest, and prioritizes the design elements constrained by the recommended regulatory clauses that are ranked higher. This determines the core direction of iterative optimization. At the same time, it dynamically adjusts the iteration step size based on the difference between the current compliance and the preset threshold. The larger the difference, the larger the iteration step size, and the smaller the difference, the smaller the iteration step size. After each iteration adjustment, the compliance of the solution is recalculated. Only the parameter adjustment results that improve the compliance are retained, while the adjustment results that do not change or decrease the compliance are discarded. During the iteration process, the compliance coefficient of all mandatory regulatory clauses is always kept at 1, while the spatial adaptability of the solution is constrained not to be lower than the preset minimum adaptability threshold, until the compliance of the solution reaches or exceeds the preset threshold, thus completing the iterative optimization of the design solution.

[0013] Furthermore, the value of the compliance discount factor follows the following: ; In the formula: Let be the compliance discount factor corresponding to the j-th recommended normative clause; Let be the control level coefficient for the j-th recommended regulatory clause; Let be the correlation coefficient between the j-th recommended specification clause and the core design objective of the target scheme; The deviation of the j-th recommended normative clause from the proposed scheme is denoted as _____.

[0014] Furthermore, after receiving a qualified design scheme, the expression module simultaneously acquires the structured design parameters output by the parsing module and the base model data generated by the construction engine. It then matches the rules and materials adapted in the preset design case knowledge base to sequentially complete the volume refinement, skin generation, and background layout. During the process, it connects with the verification module to perform compliance verification, generates a 3D urban design scene model that matches the spatial coordinate system of the base model, and then generates standardized prompt words based on the 3D scene model and structured design parameters. This drives the system's built-in AIGC engine to automatically generate and output multi-view renderings and walkthrough animations.

[0015] Furthermore, during system operation, the data management module continuously collects semantic translation process data, base model construction data, scheme iteration process data, compliance verification process data, scheme iteration process data of the generation module, and compliance verification process data of the verification module. All data are uniquely associated and bound with the output results of the expression module of the corresponding design scheme. At the same time, a unique data retrieval index is constructed using the output results of the expression module as a unique identifier.

[0016] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: This invention directly translates design requirements described in natural language into structured parameters, and uses these parameters to build a base model that is precisely matched with the urban spatial coordinate system by calling basic geographic data. This system provides a reliable geometric and spatial benchmark for subsequent design derivation, enabling design intent to be quickly realized as a calculable initial solution. Based on this, the architectural form is deduced and the scheme is iterated by using the design case knowledge base. At the same time, the standard knowledge base is connected to verify the mandatory and recommended clauses item by item in real time, so as to ensure that the scheme after each round of adjustment is always within the compliance range, reducing the risk of rework due to non-compliance with the standards in the later stage. For compliance solutions, the system further refines the blocks, generates the skin and sets up the background, and drives the built-in engine to automatically output multi-view renderings and walkthrough animations, reducing a lot of repetitive work required for manual modeling, rendering and post-processing. The system automatically links and binds the operational data of each stage of the process with the final visualized results, and uses the result file as the only search entry point, making the design conditions, adjustment records and compliance judgment results traceable. This facilitates subsequent scheme review and multi-disciplinary collaboration, thereby improving the efficiency, controllability and ease of use of the entire urban design process. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of a multimodal digital intelligent collaborative management and control and intelligent expression system for urban design; Figure 2 This is a schematic diagram of the iterative optimization and compliance verification process of the solution in this invention; Figure 3 This is a schematic diagram of the hierarchical construction structure of the module base model in this invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] The present invention will be further described below with reference to embodiments.

[0021] Example: This embodiment presents a multimodal digital intelligent collaborative management and intelligent expression system for urban design, such as... Figure 1 As shown, it includes: The parsing module is used to receive natural language interaction commands related to urban design, extract design requirements, constraints and target parameters, and perform structured translation on the extracted 3D data to output structured design parameters. After receiving natural language interaction instructions related to urban design, the parsing module performs word segmentation and semantic entity recognition on the instruction text, splitting it into design requirement entities, constraint entities, and target parameter entities. Then, it performs semantic disambiguation and hierarchical mapping on each type of entity, mapping the unstructured natural language semantics to the preset urban design full-element structured parameter spectrum. It performs structured translation on the extracted 3D data and outputs structured design parameters that correspond one-to-one with the preset urban design full-element structured parameter spectrum. Among them, the pre-defined structured parameter system for all elements of urban design: This parameter spectrum does not directly call ready-made templates, but rather, based on the conventional element division logic of urban space control in the superior management documents, it breaks down the design conditions into a tree structure layer by layer according to four dimensions: spatial positioning, development intensity, form and volume, and public facilities. The spectrum has six primary nodes, which derive more than twenty secondary parameter branches. At the end, there are quantization fields that can directly participate in numerical calculations. Each field is equipped with a value precision (area rounded to square meters, distance retained to 0.01 meters) and reasonable numerical boundaries. During the construction process, the control indicators that frequently appear in the planning and management rules of various regions were compared one by one. Parameters with the same name were merged into the same node. At the same time, empty branches were reserved for attaching local special control requirements to ensure that the translated parameter names are consistent with the terminology used by designers in their daily lives. The natural language interaction commands related to urban design are derived from the text input commands of the system's human-computer interaction interface and the standardized text commands obtained by speech recognition and translation. When the parsing module performs word segmentation and semantic entity recognition on the instruction text, it calls a pre-set corpus specific to the urban design field to perform precise word segmentation based on the domain lexicon on the received instruction text. Simultaneously, it filters out meaningless modifiers, conjunctions, and redundant modifiers in the instruction text to obtain the core semantic word segmentation units of the instruction text. Then, it performs contextual semantic association encoding on the core semantic word segmentation units through a semantic encoding network. Combined with a pre-set entity tag library for the urban design field, it performs entity classification and boundary anchoring on the encoded word segmentation units. Finally, it splits the text into design requirement entities, constraint entities, and target parameter entities. Among them, design requirement entities include entities related to land use function, design positioning, and concept orientation; constraint entities include entities related to land boundary, setback requirements, and control rules; and target parameter entities include entities related to plot ratio, building density, building height, and green space ratio. The parsing module employs a semantic coding network, specifically a context coding model based on the Transformer architecture. This network consists of 6 coding layers, a 12-head self-attention mechanism, and a hidden layer dimension of 768. The training dataset uses a natural language interaction instruction sample set from the urban design domain, with a sample size of no less than 100,000. Network training is completed using the cross-entropy loss function, with 100 iterations. The urban design domain entity label library uses a three-level labeling system. The first-level labels are divided into three categories: design requirements, constraints, and target parameters. The second-level labels are further refined to 12 categories, including land use functions, control rules, and indicator parameters. The third-level labels correspond to specific design entities. Entity boundary anchoring uses a sequence labeling algorithm based on conditional random fields, which can accurately locate and classify semantic entities. Regarding the pre-defined corpus and entity tag library specifically for the field of urban design: The original texts in the corpus are derived from design briefs, planning condition notices and project design descriptions from the public solicitation phase over the past decade. More than 30,000 valid samples have been collected. All materials have been de-identified (by removing specific project names, geographical locations and related unit information) and converted to plain text format. The text annotation work was carried out by experienced technical personnel. The text was annotated in a part-of-speech hierarchy according to the combination structure of "verb-noun-numeral". At the same time, a conversion table was compiled for synonyms (such as "retreat" and "backwards"), common abbreviations (such as "floor ratio" and "plot ratio") and frequently misspelled forms. After the annotation was completed, about 20% of each batch of samples were sampled for cross-checking. A second discussion was organized to determine the annotation of the conflicting items, so that the annotation consistency rate of each entity category was kept in the same range. When the parsing module performs semantic disambiguation and hierarchical mapping on various entities, it calculates the semantic relevance between the entity and the context based on a pre-set corpus specific to the urban design field for each type of entity obtained from the splitting process. This completes the contextual semantic disambiguation of polysemous entities and determines the unique semantic orientation of the entity in the urban design control scenario. Then, according to the hierarchical division rules of the pre-set urban design full-element structured parameter spectrum, it maps the disambiguated entities to the standardized parameter nodes of the corresponding level, completing the one-to-one correspondence between entity semantics and standardized parameters. The hierarchical division rules are set according to the three-level control dimensions of overall planning, detailed control planning, and plot urban design. Each control dimension corresponds to a set of exclusive standardized parameter nodes within the parameter spectrum. The module is used to call urban spatial basic geographic data based on structured design parameters to build an urban design base model and match spatial coordinate systems for the target plot; The structured design parameters include at least the target plot spatial boundary parameters, spatial coordinate system parameters, land use parameters, design control index parameters, and design target parameters; The construction module, based on the target plot spatial boundary parameters in the structured design parameters, calls the basic urban spatial geographic data to extract topographic elevation data, land ownership data, existing building and structure data and municipal supporting data of the target plot and its surrounding preset range, and completes the layered construction of the base model; at the same time, based on the spatial coordinate system parameters in the structured design parameters, the completed base model is subjected to coordinate system origin calibration and spatial scale normalization processing to match the base model with the overall urban spatial coordinate system; The hierarchical construction of the base model follows the following rules: S1, based on the spatial boundary parameters of the target plot, the extracted terrain elevation data is subjected to boundary clipping and elevation smoothing to generate a continuous digital elevation model covering the target plot and the surrounding preset range, thus completing the construction of the terrain benchmark layer; S2, based on the land use nature parameters in the spatial boundary parameters and structured design parameters of the target plot, performs boundary confirmation and land use attribute labeling on the extracted land ownership data, delineates the land use control boundary between the target plot and the surrounding adjacent plots, and completes the construction of the land use control layer; S3: Extract the plan outline, building height, number of floors and base coordinate information of existing buildings from the existing building data. Based on the spatial coordinate system parameters in the structured design parameters, complete the precise spatial positioning of each building's three-dimensional blocks and complete the construction of the existing building layers. S4: Extract the spatial coordinates and functional attribute information of roads, municipal pipelines, and public service facilities from the municipal infrastructure data, complete the three-dimensional spatial positioning and attribute information association and binding of various municipal infrastructure elements, and complete the construction of the municipal infrastructure layer; after the construction of each layer is completed, perform the consistency verification of spatial coordinates between layers, and after the verification is passed, complete the nesting and fusion of each layer to form a complete base model; The generation module is used to generate design concepts, deduce architectural forms, and iteratively optimize schemes based on the completed base model and structured design parameters, and call the preset design case knowledge base. The logic for constructing and filtering the pre-defined design case knowledge base: The case study database mainly consists of completed projects or projects that have received formal approval from the local planning authorities. The projects are stored in separate zones according to two independent dimensions: dominant use function and construction land scale. Each project must meet the data completeness requirements before being included in the database—the master plan, core economic and technical indicators table, typical cross-sections and spatial analysis diagrams must all be complete. After the system automatically verifies that the spatial topology is correct, the project enters the manual review queue. Manual review involves scoring the design logic coherence and spatial organization rationality of the cases. Cases that pass the score are included in the main database after completing the reverse reconstruction of the three-dimensional volume and assigning values ​​to the fields of key morphological parameters. The case studies in the database are updated on a rolling basis. At the end of each quarter, newly completed projects that meet the screening criteria are added to the database. At the same time, cases that have been in the database for more than six years and whose design concepts are significantly out of touch with the current development direction are marked as downgraded and will no longer participate in the priority matching of subsequent solutions. The generation module, based on the completed base model and structured design parameters, first uses the land use, spatial scale, and design objectives of the target plot as core matching dimensions. It then matches design cases from a pre-set design case knowledge base whose overall similarity across these dimensions meets preset thresholds. The module extracts core design concepts and architectural form control rules from these cases, and then maps the extracted elements and rules to the structured design parameters to generate the initial design scheme. Next, with spatial adaptability as the core optimization objective, it performs architectural form deduction and iterative scheme optimization to improve spatial adaptability. ; In the formula: This is the ratio of the actual floor area ratio of the proposed scheme to the target floor area ratio given in the structured design parameters. It is the minimum ratio of the actual setback distance of the building foundation in each direction to the minimum setback distance of the corresponding direction constraint; K is the ratio of the absolute value of the difference between the highest building height of the proposed scheme and the average height of existing buildings within the surrounding preset range to the preset height threshold; K is the ratio of the actual building density of the proposed scheme to the target building density given in the structural design parameters. In the formula for calculating spatial adaptability, parameter K is obtained by extracting the ratio of the total area of ​​the building base of the design scheme to the total land area of ​​the target plot, and then dividing it by the target building density. The value range is from 0 to 1. It is used to quantify the degree of matching between the building density of the scheme and the design target. Together with the plot ratio, setback ratio, and height difference ratio, it constitutes the core calculation dimension of spatial adaptability. The pre-set design case knowledge base includes case collections covering urban design projects already implemented both domestically and internationally. Cases are categorized and filtered according to land use, such as residential, commercial, and mixed-use. The structured storage of cases includes 28 core fields, such as project location, land size, design indicators, spatial layout, building form, and style characteristics. The pre-set threshold for comprehensive case similarity is set between 0.75 and 0.9 based on the land size level. The comprehensive similarity is calculated using the geometric mean of semantic matching degree of land use, numerical fitting degree of spatial scale, and matching degree of multiple elements of design objectives. The knowledge base adopts an incremental update mechanism, adding new completed project cases every quarter and completing field labeling and data entry. The above formula calculates the negative exponent after summing the squares of core parameters such as plot ratio, setback ratio, and height difference ratio. It transforms multi-dimensional spatial deviations into a unified fit value, which balances the impact of deviations of each parameter. It can accurately quantify the spatial fit level of the scheme, provide an intuitive quantitative judgment basis for the iterative optimization of the scheme, and meet the evaluation needs of simultaneous optimization of multiple parameters in urban design. During the iterative optimization process, when the spatial fit S reaches the preset threshold, the iteration stops and the optimized design solution is output. The matching criteria for design cases are as follows: the land use nature of the target plot as specified in the structured design parameters, the spatial scale corresponding to the plot's spatial boundary, and the design objectives corresponding to the design requirements and target parameters are the three core matching dimensions. The land use nature dimension uses semantic matching degree to calculate the matching value, the spatial scale dimension uses the numerical fitting degree of the plot's length, width, and area to calculate the matching value, and the design objective dimension uses multi-element semantic matching degree to calculate the matching value. The geometric mean of the matching values ​​of the three dimensions is used as the comprehensive similarity. When the comprehensive similarity is greater than or equal to the preset similarity threshold, it is determined to be a design case that meets the preset requirements. The core elements of the design concept include the spatial layout logic, functional zoning structure, style positioning guidance, and public space organization method corresponding to the case. The architectural form control rules include the building volume classification rules, building height level control rules, width-depth ratio constraint rules, building setback control rules, and spatial texture continuity rules corresponding to the case. When generating the initial design scheme, the constraints and target parameters in the structured design parameters are first decomposed into five core control parameters: land boundary constraints, plot ratio constraints, building density constraints, height control constraints, and functional allocation constraints. Then, the extracted core elements of the design concept and building form control rules are adapted to the five core control parameters. Elements and rules that conflict with the core control parameters are eliminated. Finally, based on the adjusted elements and rules, the building block layout, functional zoning, and public space layout are completed within the land boundary of the base model, generating an initial design scheme that fully conforms to the structured design parameters. In the architectural form deduction and scheme iteration optimization stage, the initial design scheme is used as the starting point for iteration, and spatial adaptability is used as the core optimization goal. A single-variable gradient descent iteration strategy is adopted to adjust the four form parameters of the building block independently one by one. After each single parameter adjustment, the spatial adaptability of the adjusted scheme is calculated simultaneously, and the adjusted scheme is pushed to the verification module to obtain the real-time compliance verification result. If the spatial adaptability of the adjusted scheme is improved compared with the previous iteration and the compliance verification result meets the preset requirements, the parameter adjustment is retained. Otherwise, the adjustment is rolled back and the iteration step size is reduced. The above iteration process is repeated until the spatial adaptability reaches the preset threshold or the number of iterations reaches the preset maximum number of iterations, and the final optimized design scheme is output. The spatial adaptability preset threshold for scheme iteration optimization is set to 0.85 to 0.95 according to the land parcel type, and the standard compliance preset threshold is uniformly set to 0.9. The minimum iteration step size is divided into three levels: 0.1 meters, 0.01, and 1% according to the design parameter type. The iteration step size is dynamically adjusted according to the difference between the current compliance and the target threshold. The maximum number of iterations is set to 50 to 100 rounds according to the scheme complexity. During the iteration process, the parameter adjustment results of compliance and improved adaptability are retained in real time to ensure the efficiency and accuracy of scheme optimization. The verification module is used to connect to the preset urban design code knowledge base, perform real-time verification of all elements of the design scheme generated iteratively, output the verification results and synchronize them to the generation module; The pre-set urban design code knowledge base is formed by first collecting currently valid legal code documents related to urban design at the national, industry, and local levels, breaking down and extracting individual executable code clauses, matching each code clause with the corresponding applicable land parcel type, constraint control elements, compliance judgment rules, and boundary parameter thresholds, and then constructing the knowledge base after completing the hierarchical classification and structured storage of all code clauses. The urban design code knowledge base is divided into independent control rules for each code clause. Mandatory codes and recommended codes are classified according to the validity of the code text. Applicable land types and constraint control elements are automatically associated using a keyword matching algorithm. Boundary parameter thresholds are extracted and stored from the code clauses. Compliance judgment uses three algorithms: numerical comparison, spatial relationship verification, and index matching. It can automatically complete the full-element comparison between the design scheme and the code clauses and accurately output the compliance coefficient and code compliance. After the verification module connects to the pre-set urban design code knowledge base, it first categorizes the code clauses in the knowledge base into two types: mandatory code clauses and recommended code clauses. Then, for each iteratively generated design scheme, it performs a compliance determination for each of the two types of clauses, and outputs the code compliance score of the scheme. ; In the formula: This refers to the total number of mandatory regulatory clauses in the pre-defined urban design code knowledge base that match the target plot. This refers to the total number of recommended code clauses in the pre-defined urban design code knowledge base that match the target plot. For the first The compliance coefficient of a mandatory regulatory clause is 1 when the solution meets the requirements of the clause and 0 when it does not. For the first The compliance coefficient of the recommended normative clause is 1 when the solution meets the requirements of the clause, and 1 when it does not meet the requirements. This formula uses a combination of the product of mandatory normative clauses and the geometric mean of recommended normative clauses to distinguish the control priorities of the two types of norms, accurately calculate the overall compliance of the scheme, clearly reflect the scheme's satisfaction with norms of different effects, and provide a scientific quantitative reference for the scheme's optimization direction and step size adjustment. The verification module will standardize compliance. The calculation results are synchronized to the generation module as verification results, and the generation module adjusts the direction and step size of the scheme iteration optimization based on the verification results; After receiving the verification results, the generation module determines the compliance status of the solution based on the calculation results of the compliance degree. When the compliance coefficient of any mandatory standard clause is 0, the generation module locks the design element constrained by the corresponding non-compliant mandatory standard clause. Taking the elimination of the non-compliance of the mandatory standard clause as the only direction of iterative optimization, the iteration step size is adjusted to the preset minimum step size. Simultaneously, the boundary parameter threshold corresponding to the non-compliant mandatory standard clause is extracted to determine the compliance value range of the locked design element. The parameter value of the design element in the current solution is used as the adjustment starting point, and the preset minimum step size is used as the single adjustment unit. The adjustment is carried out unidirectionally and successively within the compliance value range. After each adjustment, only the compliance verification of the mandatory standard clause is re-performed. During the adjustment process, other non-locked compliant design elements are not modified until the compliance coefficient of the mandatory standard clause changes to 1, so as to complete the correction of all non-compliant mandatory standard clauses. When the compliance coefficient of all mandatory regulatory clauses is 1, the generation module takes improving the compliance of the regulations to a preset threshold as its core objective. It obtains all unmet recommended regulatory clauses, sorts them by their corresponding compliance discount coefficients from smallest to largest, and prioritizes the design elements constrained by the recommended regulatory clauses that are ranked first. This determines the core direction of iterative optimization. At the same time, it dynamically adjusts the iteration step size based on the difference between the current compliance of the regulations and the preset threshold. The larger the difference, the larger the iteration step size, and the smaller the difference, the smaller the iteration step size. After each iteration adjustment, the compliance of the scheme is recalculated. Only the parameter adjustment results that improve the compliance of the regulations are retained, while the adjustment results that do not change or decrease the compliance of the regulations are discarded. During the iteration process, the compliance coefficient of all mandatory regulatory clauses is always kept at 1, while the spatial adaptability of the scheme is constrained not to be lower than the preset minimum adaptability threshold, until the compliance of the scheme reaches or exceeds the preset threshold, thus completing the iterative optimization of the design scheme. The value of the compliant discount factor follows the following rules: ; In the formula: Let be the compliance discount coefficient corresponding to the j-th recommended normative clause, with a value ranging from 0 to 1; Let be the control level coefficient for the j-th recommended regulatory clause, with a value ranging from 0 to 1. This coefficient is preset according to the effectiveness level of the regulation to which the clause belongs; the higher the effectiveness level, the better. The closer the value is to 1; Let be the correlation coefficient between the j-th recommended specification clause and the core design objective of the target scheme. The value range is from 0 to 1. It is calculated based on the matching degree between the target parameters in the structured design parameters and the content controlled by the clause. The higher the matching degree, the better. The closer the value is to 1; The deviation of the j-th recommended specification clause is denoted by 0, and its value range is from 0 to 1. It is the ratio of the absolute value of the difference between the actual parameters of the scheme and the parameters required by the clause to the maximum allowable fluctuation threshold preset by the clause. The above formula combines three types of factors—control level, target relevance, and plan deviation—by multiplying them and taking the negative exponent. This transforms the multi-dimensional influencing factors into a single discount coefficient, accurately quantifying the degree of compliance loss of recommended norms. The expression module is used to receive the design scheme that has passed the verification, and simultaneously perform block refinement, skin generation, and background layout on the design scheme, and drive the system's built-in AIGC engine to automatically generate renderings and walkthrough animations. After receiving the approved design scheme, the expression module synchronously obtains the structured design parameters output by the parsing module and the base model data generated by the construction engine. It then matches the rules and materials in the preset design case knowledge base to complete the volume refinement, skin generation and background layout in sequence. During the process, it connects with the verification module to perform compliance verification, generates a three-dimensional urban design scene model that matches the spatial coordinate system of the base model, and then generates standardized prompt words based on the three-dimensional scene model and structured design parameters. This drives the system's built-in AIGC engine to automatically generate and output multi-view renderings and walkthrough animations. Among them, the block refinement takes the land use control requirements and design control indicators in the structured design parameters as rigid constraints, matches the refinement rules from the preset design case knowledge base, and completes the planar outline correction, vertical functional hierarchy division and internal space alignment mapping of the building block. During the process, each parameter adjustment is pushed to the verification module, which calls the mandatory normative clauses of the urban design code knowledge base to verify the core indicators of building setback, height, density and plot ratio in real time. Only the adjustment results with a full compliance coefficient of 1 are retained, and the building block model that matches the coordinate system of the base model is output. During the skin generation stage, the corresponding skin material library and facade construction rules are matched from the preset design case knowledge base. According to the building function constraints, the facade windows, texture and material selection are generated to generate a skin model that is bound to the spatial coordinates of the block. The verification module is pushed simultaneously to complete the compliance verification of the style and facade control clauses and then locks the model. When setting up the background, the parameters of public space, roads and green space in the target plot and the surrounding preset range are extracted. The matching elements are matched from the preset background material library to complete the automatic placement of road markings, landscape greenery, supporting facilities and site paving. During the process, the verification module is pushed to complete the spatial collision verification and the compliance verification of public space and green space ratio related specifications. After eliminating the conflict, the complete urban design 3D scene model is output. The background material library integrates authorized general 3D model resources and reusable component models that have been verified in self-owned projects, and is divided into four categories according to application scenarios: landscape vegetation, street furniture, traffic facilities and site paving. Before each model is added to the database, it must undergo three standardized processing steps: polygon reduction and compression (controlling the number of polygons within an appropriate range), material texture format standardization, and collision bounding box generation. After processing, each model is placed into a standard-sized test scene for scale and color temperature comparison, and models with excessive deviation from the actual space ratio or obvious color distortion are eliminated. The spatial placement rules required for scene layout (such as the spacing between street trees, the spacing between streetlights, the orientation of seats, etc.) are not bound to the model itself, but are mounted in the management directory of the material library in the form of an independent configuration table, which makes it easy to make batch adjustments for different street scales; Standardized prompt words are generated by extracting the spatial layout, architectural style, and scale parameters of the 3D scene model, combining the design concept, positioning, and control indicators of the structured design parameters, and breaking them down into four semantic units according to the AIGC engine input specifications: scene subject, style limitation, parameter constraint, and image quality requirements. After being standardized, standardized prompt words that conform to the input format are generated. The built-in AIGC engine is a 3D rendering and generation engine adapted to urban design scenarios. It has a built-in dedicated rendering model library, view parameter library and video encapsulation unit. After receiving prompt words and 3D models, the AIGC engine renders and generates multi-view static effect images according to preset parameters. At the same time, it generates a collision-free roaming path along the spatial axis of the target plot, the sequence of public spaces and the core display surface of the building, and completes the frame rendering, audio-visual matching and standardized format encapsulation output of the roaming animation. The built-in AIGC engine uses the StableDiffusion 3D rendering base model to make specific fine-tuning of architectural style, site landscape and municipal facilities for urban design scenarios. The rendering parameters are uniformly set to a resolution of 3840×2160, a rendering accuracy of 4K and a lighting mode of urban natural light. The standardized prompts use a fixed decomposition template, which is divided into four modules: scene subject, design style, indicator constraints and image quality requirements. The roaming path generation uses the A* collision detection algorithm to plan the path along the spatial axis of the plot and the sequence of public spaces. The frame rendering parameters are set to 30 frames / second, and the animation encapsulation and output are completed automatically. The management module is used to associate and bind the runtime data of the parsing module, building module, generating module, and verification module with the output results of the expression module for storage, and to set the output results of the expression module as a unique data retrieval index. During system operation, the data management module continuously collects data from the semantic translation process, the base model construction process, the solution iteration process, the compliance verification process, the solution iteration process of the generation module, and the compliance verification process of the verification module. All data are uniquely associated and bound with the output results of the expression module of the corresponding design solution. At the same time, a unique data retrieval index is constructed using the output results of the expression module as a unique identifier. The parsing module interacts with the building module via a wireless network. The building module interacts with the generation module via a wireless network. The generation module interacts with the verification module via a wireless network. The verification module interacts with the expression module and the management module via a wireless network. It should be noted that: All numerical preset parameters in the system are not fixed constants, but rather empirical value ranges determined based on the statistical distribution of indicators from a large number of approved historical schemes and the convergence results of multiple rounds of actual measurements from the prototype system. For example, the iteration stop value of spatial adaptability is set to a slightly gradient value according to different land use functions—residential plots focus on setbacks and sunlight, while commercial plots focus on line coverage and height display, so the convergence threshold is slightly raised. The compliance threshold is based on the prevailing expectations in approval practice regarding the overall implementation ratio of various flexible clauses; The initial value of the iteration step size is determined based on the commonly used precision units of each parameter in actual measurement. After trial runs of dozens of different sets of calculation examples, it is adjusted to a balanced state that can avoid the number of iterations from being too small and avoid the step size from being too large and skipping the compliant narrow area. All adjustable thresholds are not fixed values ​​in the system. After deployment, users can adapt to any threshold by scaling it up or fine-tuning it individually.

[0022] In this embodiment, after the system starts, the text or voice commands input by the user first enter the parsing module. The module first performs word segmentation and semantic recognition on the commands, breaks down the design requirements, constraints and target parameters, and then translates them into a set of structured design parameters through disambiguation and hierarchical mapping. These parameters include plot boundaries, coordinate system, land use nature and control indicators such as plot ratio and building density. This process transforms the ambiguous natural language into standard data that the machine can use directly, avoiding common deviations in human understanding and unifying the language for all subsequent stages. Structured parameters flow to the building module, which retrieves data such as topography, ownership, existing buildings and municipal facilities from the city's basic geographic database based on the land parcel boundaries. The base model is built layer by layer. After each layer is built, the system automatically performs spatial coordinate consistency verification, then coordinate system calibration and scale normalization, so that the base model is completely consistent with the city's overall coordinate system. In this way, multi-source data with different sources and inconsistent coordinates are integrated under the same benchmark, and subsequent design and verification have a reliable spatial base map. After acquiring the base model and design parameters, the generation module first matches highly similar design cases in the case knowledge base, extracts the design concepts and form control rules, and then adapts and adjusts these rules and parameters to generate an initial scheme within the land boundary. The module then iteratively optimizes the scheme based on spatial adaptability, adjusting the position, height, volume, or setback of the blocks in each round, and then evaluating whether the adaptability has improved. At the same time, the scheme is sent to the verification module for compliance checks. Only adjustments that improve adaptability and meet the specifications will be retained; otherwise, they will be rolled back and the step size will be reduced. This process is repeated until the adaptability meets the standard. This automatic optimization process avoids the work of designers repeatedly calculating, and each round of adjustment is bidirectionally verified, with a clear and reliable convergence direction. While the design is iterating, the verification module runs synchronously in the background. It connects to the established specification knowledge base, divides specification clauses into mandatory and recommended categories, and judges the design scheme clause by clause. If a mandatory clause is violated, it is directly judged as non-compliant. For recommended clauses, a discount coefficient is given according to the degree of deviation. The compliance degree of the scheme is calculated in a comprehensive manner. This compliance degree result is fed back to the generation module in real time. When a mandatory violation occurs, the generation module will lock the corresponding design element and adjust it unidirectionally towards the compliance range in the smallest step, without changing other elements, until the violation is eliminated. After all mandatory violations are passed, the recommended clauses that have the greatest impact on compliance degree are adjusted first, and the overall compliance level is steadily improved. This "design and verify at the same time" model effectively changes the traditional post-approval process and eliminates the need for rework after the design is finalized from the source. After the design scheme passes verification, it enters the expression module. The expression module retrieves the structured design parameters and base model, and then performs volume refinement, skin generation, and background layout in sequence. Each step is verified as it is done, ultimately forming a complete 3D scene model of the urban design. Then, the system generates standardized prompts based on the spatial layout and design parameters of the model, driving the built-in AIGC engine to automatically render multi-view renderings and walkthrough animations. In this way, professional visualization results can be obtained at the same time as the design scheme is finalized, without the need to separately commission a rendering team to remodel, and the delivery cycle is greatly shortened. Meanwhile, the management module records the process data of each stage of parsing, building, generating and verifying in the background in real time, and binds all data to the final generated renderings, animations and 3D models. The visualized results serve as the sole retrieval index. In the future, as long as the result file is viewed, all process data of this plan from the original input to the iteration records and then to the verification report can be retrieved with one click. The entire system connects urban design from demand input, modeling, plan generation, compliance verification, result expression to data archiving into a complete pipeline. All operations flow automatically in a closed loop, which not only improves efficiency, but also ensures the compliance and traceability of the plan.

[0023] See Figure 2 As shown in the figure, the complete processing flow from top to bottom includes design case matching, initial scheme generation, spatial adaptability calculation, iterative optimization and adjustment, compliance verification, standard compliance calculation, and achievement output. The compliance verification process covers two types of verification paths: mandatory standard judgment and recommended standard judgment. The verification results output from the standard compliance calculation process are fed back to the spatial adaptability calculation process through an iterative loop path, thus forming a closed-loop optimization mechanism for bidirectional collaboration between the generation module and the verification module.

[0024] See Figure 3 As shown in the figure, the diagram illustrates a four-layer stacked structure from top to bottom: S1 Municipal Supporting Facilities Layer, S2 Existing Buildings and Structures Layer, S3 Land Use Control Layer, and S4 Topographic Baseline Layer. These layers correspond to the construction of four levels: three-dimensional positioning of road networks and public service facilities, spatial layout of existing building blocks, confirmation of land ownership and delineation of control boundaries, and cropping and smoothing of topographic elevation data. After the four-layer structure undergoes coordinate consistency verification and nested fusion processing, it forms a complete base model that matches the overall urban spatial coordinate system, thus demonstrating the layered composition architecture and fusion construction method of the base model.

[0025] In summary, the system in the above embodiments directly translates design requirements described in natural language into structured parameters, and uses these parameters to build a base model that precisely matches the urban spatial coordinate system by calling basic geographic data. This provides a reliable geometric and spatial benchmark for subsequent design derivation, enabling design intent to be quickly translated into calculable initial solutions. Based on this, the system guides architectural form derivation and solution iteration using a design case knowledge base, while simultaneously accessing a regulatory knowledge base to perform real-time verification of mandatory and recommended clauses, ensuring that each adjusted solution remains within compliance limits. This reduces the risk of rework due to regulatory non-compliance. For compliant solutions, the system further refines the massing, generates the skin, and sets up the background, and drives the built-in engine to automatically output multi-view renderings and walkthrough animations, reducing the large amount of repetitive work required for manual modeling, rendering, and post-processing. Throughout the entire process, the system's operational data and the final visualization results are automatically linked and bound, with the result file serving as the sole retrieval entry point. This ensures that design conditions, adjustment records, and compliance judgment results are traceable, facilitating subsequent solution review and multi-disciplinary collaboration. As a result, the system improves the efficiency, controllability, and ease of use of the entire urban design process.

[0026] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 will 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 multimodal digital intelligent collaborative control and intelligent expression system for urban design, characterized in that, include: The parsing module is used to receive natural language interaction commands related to urban design, extract design requirements, constraints and target parameters, and perform structured translation on the extracted 3D data to output structured design parameters. The module is used to call urban spatial basic geographic data based on structured design parameters to build an urban design base model and match spatial coordinate systems for the target plot; The generation module is used to generate design concepts, deduce architectural forms, and iteratively optimize schemes based on the completed base model and structured design parameters, and call the preset design case knowledge base. The verification module is used to connect to the preset urban design code knowledge base, perform real-time verification of all elements of the design scheme generated iteratively, output the verification results and synchronize them to the generation module; The expression module is used to receive the design scheme that has passed the verification, and simultaneously perform block refinement, skin generation, and background layout on the design scheme, and drive the system's built-in AIGC engine to automatically generate renderings and walkthrough animations. The management module is used to associate and bind the full-process running data of the parsing module, construction module, generation module, and verification module with the output results of the expression module for storage, and to set the output results of the expression module as a unique data retrieval index. After receiving the verification results, the generation module determines the compliance status of the solution based on the calculation results of the compliance degree. When the compliance coefficient of any mandatory standard clause is 0, the generation module locks the design element constrained by the corresponding non-compliant mandatory standard clause. Taking the elimination of the non-compliance of the mandatory standard clause as the only direction of iterative optimization, the iteration step size is adjusted to the preset minimum step size. Simultaneously, the boundary parameter threshold corresponding to the non-compliant mandatory standard clause is extracted to determine the compliance value range of the locked design element. The parameter value of the design element in the current solution is used as the adjustment starting point, and the preset minimum step size is used as the single adjustment unit. The adjustment is carried out unidirectionally and successively within the compliance value range. After each adjustment, only the compliance verification of the mandatory standard clause is re-performed. During the adjustment process, other non-locked compliant design elements are not modified until the compliance coefficient of the mandatory standard clause changes to 1, so as to complete the correction of all non-compliant mandatory standard clauses. When the compliance coefficient of all mandatory regulatory clauses is 1, the generation module takes improving the compliance to a preset threshold as its core objective. It obtains all unmet recommended regulatory clauses, sorts them by their corresponding compliance discount coefficients from smallest to largest, and prioritizes the design elements constrained by the recommended regulatory clauses that are ranked higher. This determines the core direction of iterative optimization. At the same time, it dynamically adjusts the iteration step size based on the difference between the current compliance and the preset threshold. The larger the difference, the larger the iteration step size, and the smaller the difference, the smaller the iteration step size. After each iteration adjustment, the compliance of the solution is recalculated. Only the parameter adjustment results that improve the compliance are retained, while the adjustment results that do not change or decrease the compliance are discarded. During the iteration process, the compliance coefficient of all mandatory regulatory clauses is always kept at 1, while the spatial adaptability of the solution is constrained not to be lower than the preset minimum adaptability threshold, until the compliance of the solution reaches or exceeds the preset threshold, thus completing the iterative optimization of the design solution.

2. The multimodal digital intelligent collaborative management and intelligent expression system for urban design according to claim 1, characterized in that, After receiving natural language interaction instructions related to urban design, the parsing module performs word segmentation and semantic entity recognition on the instruction text, splitting it into design requirement entities, constraint entities, and target parameter entities. Then, it performs semantic disambiguation and hierarchical mapping on each type of entity, mapping the unstructured natural language semantics to the preset urban design full-element structured parameter spectrum. The extracted three-dimensional data is then translated into structure, and the structured design parameters that correspond one-to-one with the preset urban design full-element structured parameter spectrum are output.

3. The multimodal digital intelligent collaborative management and intelligent expression system for urban design according to claim 1, characterized in that, The structured design parameters include at least the target plot spatial boundary parameters, spatial coordinate system parameters, land use parameters, design control index parameters, and design target parameters; The construction module, based on the target plot spatial boundary parameters in the structured design parameters, calls urban spatial basic geographic data to extract topographic elevation data, land ownership data, existing building and structure data and municipal supporting data of the target plot and its surrounding preset range, and completes the layered construction of the base model; at the same time, based on the spatial coordinate system parameters in the structured design parameters, the completed base model is subjected to coordinate system origin calibration and spatial scale normalization processing to match the base model with the overall urban spatial coordinate system.

4. The multimodal digital intelligent collaborative management and intelligent expression system for urban design according to claim 1, characterized in that, The generation module, based on the completed base model and structured design parameters, first uses the land use nature, spatial scale, and design objectives of the target plot as the core matching dimensions. It then matches design cases from the preset design case knowledge base whose comprehensive similarity in each dimension meets the preset threshold, extracts the core elements of the design concept and the building form control rules from the cases, and then performs adaptive mapping between the extracted elements and rules and the structured design parameters to complete the generation of the initial design scheme. Then, with spatial adaptability as the core optimization objective, architectural form deduction and iterative optimization of the scheme are performed. The aforementioned spatial adaptability... ; In the formula: This is the ratio of the actual floor area ratio of the proposed scheme to the target floor area ratio given in the structured design parameters. It is the minimum ratio of the actual setback distance of the building foundation in each direction to the minimum setback distance of the corresponding direction constraint; K is the ratio of the absolute value of the difference between the highest building height of the proposed scheme and the average height of existing buildings within the surrounding preset range to the preset height threshold; K is the ratio of the actual building density of the proposed scheme to the target building density given in the structural design parameters. During the iterative optimization process, when the spatial fit S reaches the preset threshold, the iteration stops and the optimized design solution is output.

5. A multimodal digital intelligent collaborative control and intelligent expression system for urban design according to claim 1, characterized in that, After the verification module connects to a pre-set urban design code knowledge base, it first divides the code clauses in the knowledge base into two categories: mandatory code clauses and recommended code clauses. Then, for each iteratively generated design scheme, it performs a compliance determination on each of the two types of clauses, and outputs the code compliance score of the scheme. ; In the formula: This refers to the total number of mandatory regulatory clauses in the pre-defined urban design code knowledge base that match the target plot. This refers to the total number of recommended code clauses in the pre-defined urban design code knowledge base that match the target plot. For the first The compliance coefficient of each mandatory regulatory clause; For the first The compliance coefficient of the recommended normative clause is 1 when the solution meets the requirements of the clause, and 1 when it does not meet the requirements. The verification module will standardize compliance. The calculation results are synchronized to the generation module as verification results. The generation module adjusts the direction and step size of the scheme iteration optimization based on the verification results.

6. The multimodal digital intelligent collaborative control and intelligent expression system for urban design according to claim 5, characterized in that, The value of the compliance discount factor follows the following rules: ； In the formula: Let be the compliance discount factor corresponding to the j-th recommended normative clause; Let be the control level coefficient for the j-th recommended regulatory clause; Let be the correlation coefficient between the j-th recommended specification clause and the core design objective of the target scheme; The deviation of the j-th recommended normative clause from the proposed scheme is denoted as _____.

7. A multimodal digital intelligent collaborative control and intelligent expression system for urban design according to claim 1, characterized in that, After receiving a qualified design scheme, the expression module synchronously obtains the structured design parameters output by the parsing module and the base model data generated by the construction module. It then matches the rules and materials in the preset design case knowledge base to complete the volume refinement, skin generation, and background layout in sequence. During the process, it connects with the verification module to perform compliance verification, generates a 3D urban design scene model that matches the spatial coordinate system of the base model, and then generates standardized prompt words based on the 3D scene model and structured design parameters. This drives the system's built-in AIGC engine to automatically generate and output multi-view renderings and walkthrough animations.

8. A multimodal digital intelligent collaborative control and intelligent expression system for urban design according to claim 1, characterized in that, During system operation, the management module continuously collects data from the semantic translation process, the base model construction process, the solution iteration process, and the compliance verification process. It then uniquely associates and binds all data with the output results of the expression module of the corresponding design solution, and uses the output results of the expression module as a unique identifier to construct a unique data retrieval index.

9. A multimodal digital intelligent collaborative control and intelligent expression system for urban design according to claim 1, characterized in that, The parsing module is interconnected with the construction module via a wireless network. The construction module is interconnected with the generation module via a wireless network. The generation module is interconnected with the verification module via a wireless network. The verification module is interconnected with the expression module and the management module via a wireless network.

Images