Multi-scene simulation deduction method based on territorial space planning

Abstract

The invention relates to the technical field of territorial space planning, in particular to a multi-scenario simulation deduction method based on territorial space planning, which comprises the following steps: collecting territorial space information data and initial territorial space planning, and constructing an intelligent agent representing planners of multiple parties according to the initial space planning; a speaking rule based on an attention mechanism is adopted to drive a multi-party agent to speak for multiple rounds for the initial territorial plan; through a dynamic game algorithm, the multi-party agents are promoted to continuously adjust the initial reconstruction territorial space planning in multiple rounds of interactive speaking, and a quadratic planning scheme is formed; according to the method, a plurality of agents are constructed and form a perfect space planning scheme based on a dynamic game algorithm, so that the problem that a conventional land space planning method is difficult to implement when collecting and integrating opinions of a plurality of persons is solved; and the communication efficiency is low due to the fact that the method mainly depends on offline conference discussion, paper questionnaire investigation and the like.

Description

Technical Field

[0001] This invention relates to the field of land and space planning technology, and more particularly to a multi-scenario simulation and deduction method based on land and space planning. Background Technology

[0002] Territorial spatial planning is a complex and highly specialized system engineering project that requires the collaborative participation of urban planners, geographers, ecological experts, economic analysts, and other planning professionals. Traditional territorial spatial planning relies primarily on offline meetings and paper-based questionnaires to collect and integrate opinions from various stakeholders. This communication model is inefficient, prone to delays in information delivery and misunderstandings. Furthermore, due to the differing professional backgrounds and focuses of various stakeholders, information is easily scattered and lacks systematic integration during the exchange of opinions, making it difficult to formulate a comprehensive and feasible planning solution.

[0003] In the application of multi-scenario simulation and deduction technology to territorial spatial planning, existing methods have significant shortcomings in integrating the opinions of multiple planning personnel. Most simulation and deduction models operate within a predetermined framework, lacking flexible interfaces and interactive mechanisms, and failing to effectively integrate the unique insights and personalized needs of different planners from their professional perspectives into the model parameters and scenario settings. Furthermore, different planners interpret and evaluate simulation results differently, and existing simulation and deduction technologies lack a unified visual interactive platform, making it difficult to achieve real-time collaborative analysis and dynamic feedback adjustments of simulation results from multiple parties. This leads to a disconnect between simulation and deduction results and actual planning needs, failing to fully leverage the supporting role of the wisdom of multiple planning personnel in planning decisions. Therefore, researching a multi-scenario simulation and deduction method for territorial spatial planning that can efficiently integrate the opinions of multiple planning personnel has become an urgent need to improve the scientific rigor and practicality of planning. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides a multi-scenario simulation and deduction method based on territorial spatial planning. This method can quickly integrate opinions from multiple parties and effectively incorporate these opinions into model parameters and scenario settings, thereby enabling the analysis of multi-scenario simulation results for territorial spatial planning.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A multi-scenario simulation and deduction method based on territorial spatial planning includes:

[0007] Collect land and space information data and initial land and space plans, and construct an intelligent agent representing multiple planning personnel based on the initial spatial plans;

[0008] An attention-based speaking rule is adopted to drive multiple agents to speak in multiple rounds regarding the initial land use plan;

[0009] Through dynamic game theory algorithms, multiple intelligent agents are prompted to continuously adjust the initial land space transformation plan through multiple rounds of interactive discussions, thus forming a secondary planning scheme.

[0010] A holographic model is constructed based on a quadratic programming scheme, and multi-scenario simulations are performed based on the holographic model.

[0011] Preferably, before the process of driving the multi-party intelligent agents to make multiple rounds of statements regarding the initial land use plan, it also includes:

[0012] Calculate the Pearson correlation coefficient among the multi-agent agents using historical interaction data from the database.

[0013] The correlation between agents is determined based on the value of the Pearson correlation coefficient;

[0014] Different dynamic game algorithms are used based on the relationships between agents.

[0015] Preferably, after constructing the holographic model based on the quadratic programming scheme and performing multi-scenario simulations based on the holographic model, the method further includes:

[0016] The agent obtains the simulation results and determines whether to adopt a secondary planning scheme based on the simulation results.

[0017] If so, adopt a secondary planning scheme and use the secondary planning scheme as the final planning scheme;

[0018] If not, it will drive multiple intelligent agents to continue speaking on the secondary planning scheme in multiple rounds until a final planning scheme is formed.

[0019] Preferably, the construction of the holographic model based on the quadratic programming scheme, and the multi-scenario simulation based on the holographic model, include:

[0020] The secondary planning scheme is input into the BIM and GIS simulation platform, a three-dimensional model is constructed through the BIM and GIS simulation platform, and a holographic planning model is constructed based on the three-dimensional model.

[0021] Import the holographic planning model into a predetermined scenario, and perform dynamic simulation and deduction based on the predetermined scenario;

[0022] The simulation results are analyzed from multiple dimensions, and the results are then visualized.

[0023] Preferably, before adopting the attention-based speaking rules to drive multiple agents to speak in multiple rounds regarding the initial land use plan, the following further steps are included:

[0024] Obtain the influencing factors of land and space planning, and map these influencing factors to multi-party intelligent agents;

[0025] Based on the weights of the influencing factors, the speaking weights of the multi-party agents are assigned, and different speaking times are assigned to the multi-party agents based on their speaking weights.

[0026] Preferably, before obtaining the influencing factors of territorial spatial planning, the method further includes:

[0027] Based on a literature review and policy documents, a pool of potential influencing factors was compiled.

[0028] The contribution of each factor in the influencing factor pool is calculated using GA and machine learning models.

[0029] Based on the contribution values ​​of each factor, list several major influencing factors;

[0030] GWR analysis was used to analyze the regional differences in the influence of major influencing factors and to identify universal key factors and local dominant factors.

[0031] The key factors and local dominant factors were submitted to experts for review, and their weights were adjusted or qualitative constraints were added.

[0032] Establish a feedback mechanism to periodically reassess the weights of influencing factors based on new data.

[0033] Preferably, the construction of the holographic planning model based on the three-dimensional model includes:

[0034] The 3D model is converted and optimized in holographic format to obtain the converted holographic model framework;

[0035] Determine the planning entity elements of the holographic model framework, the optical attribute information of each type of planning entity element, and the geometric parameter information;

[0036] Based on the geometric parameters, construct parametric geometric models corresponding to various planning entity elements;

[0037] Obtain the texture map of the parametric geometry model and map the texture map onto the parametric geometry;

[0038] The aforementioned geometric parameters are imported into the holographic model framework to generate an initial holographic planning model.

[0039] Based on the initial holographic planning model, and combined with the optical attribute information of each type of entity element, a shader is used for lighting rendering to generate a holographic planning model with realistic lighting effects.

[0040] Preferably, the generation of the initial spatial plan includes:

[0041] Integrate multi-source data and preprocess the multi-source data;

[0042] Obtain preprocessed multi-source data and mine spatial patterns from the multi-source data using association rule algorithms;

[0043] By using a convolutional neural network to combine historical successful planning schemes in the database with spatial patterns from multiple sources, an initial spatial plan that conforms to the specifications is generated.

[0044] Preferably, the adoption of attention-based speaking rules to drive multiple agents to speak in multiple rounds regarding the initial land use plan includes:

[0045] During multiple rounds of speaking, the speaking data of the various intelligent agents is collected in real time through monitoring equipment;

[0046] The speech data is analyzed and detected using detection models and algorithms;

[0047] Based on the analysis and detection results, dynamic intervention is carried out on the non-compliant intelligent agents.

[0048] Preferably, the multi-dimensional analysis of the simulation results includes economic, ecological, social maintenance, and policy risk dimensions.

[0049] The beneficial effects of this invention are as follows:

[0050] 1. This invention constructs multiple intelligent agents, which form a complete spatial planning scheme based on a dynamic game algorithm. This solves the problem that traditional land spatial planning mainly relies on offline meetings and paper questionnaires to collect and integrate opinions from multiple parties, resulting in low communication efficiency and problems such as untimely information transmission and misunderstandings.

[0051] 2. By presenting spatial planning schemes in the form of three-dimensional and holographic models, scenario simulation can achieve three-dimensional visualization, dynamic interaction, immersive experience, and multi-source data fusion, significantly improving the scientific nature of planning and the accuracy of simulation. Attached Figure Description

[0052] Figure 1 This is a flowchart of a multi-scenario simulation and deduction method based on territorial spatial planning in a specific embodiment of the present invention;

[0053] Figure 2 This is a flowchart illustrating the selection process of the dynamic game algorithm in a specific embodiment of the present invention;

[0054] Figure 3 This is a flowchart illustrating multi-scenario simulation and deduction based on a holographic model in a specific embodiment of the present invention;

[0055] Figure 4This is a flowchart illustrating the setting of different weight values ​​for different agents in a specific embodiment of the present invention;

[0056] Figure 5 This is a flowchart illustrating the construction of a holographic planning model in a specific embodiment of the present invention;

[0057] Figure 6 This is a flowchart illustrating the initial space planning generation process in a specific embodiment of the present invention;

[0058] Figure 7 This is a flowchart illustrating the monitoring of agent speech in a specific embodiment of the present invention;

[0059] Figure 8 This is a preset scenario table in a specific embodiment of the present invention. Detailed Implementation

[0060] Example 1

[0061] Please see Figures 1-8 As shown, this invention relates to a multi-scenario simulation and deduction method based on territorial spatial planning.

[0062] An intelligent agent is a software or hardware system that can perceive its environment, make decisions based on perceived information, and execute actions to achieve specific goals. It can operate completely autonomously or under the supervision of a human operator. Intelligent agents use built-in algorithms to analyze data, formulate plans, and take actions to complete predetermined tasks.

[0063] Specific deduction methods include:

[0064] Acquire preprocessed multi-source data and perform preprocessing on the multi-source data, including data evaluation, data standardization and data cleaning, and mine the spatial patterns of multi-source data through association rule algorithms;

[0065] By using a convolutional neural network to combine historical successful planning schemes in the database with spatial patterns from multiple sources, an initial spatial plan that conforms to the specifications is generated.

[0066] Collect land and space information data and initial land and space plans, and construct an intelligent agent representing multiple planning parties based on the initial spatial plan. Specifically, the representatives of the intelligent agent are: government and functional departments, technical expert teams, representatives of the public, development enterprises, and the National People's Congress and the Chinese People's Political Consultative Conference.

[0067] Specifically, spatial information data includes landform type, shape, area, topography, water distribution, land location coordinates, remote sensing images of the land, and land type.

[0068] Before the multi-agent speech, the Pearson correlation coefficient between the multi-agents is calculated using the historical interaction data of the multi-agents in the database.

[0069] The correlation between agents is determined by the value of the Pearson correlation coefficient, which is denoted as ρ. i，j ;

[0070] When ρ i，j A correlation of >0.5 indicates a strong positive correlation, suggesting a high probability of cooperation among related agents (e.g., mutual benefit in collaborative tasks).

[0071] When ρ i，j A correlation of less than -0.5 indicates a strong negative correlation, suggesting a high probability of competition (e.g., resource competition in a zero-sum game).

[0072] When -0.5≤ρ i，j ≤0.5 indicates a weak correlation or a mixed relationship.

[0073] Different dynamic game algorithms are used based on the relationships between agents.

[0074] The identification adopts attention-based speaking rules to drive multiple agents to speak in multiple rounds regarding the initial land use plan;

[0075] The attention mechanism's speaking rules, which dynamically allocate weights to focus on key information, require collecting factors influencing the national spatial planning before driving multiple agents to speak on the initial land use plan. These factors include:

[0076] Based on a literature review and policy documents, a pool of potential influencing factors was compiled.

[0077] The contribution of each factor in the influencing factor pool is calculated using GA and machine learning models.

[0078] Based on the contribution values ​​of each factor, list several major influencing factors;

[0079] GWR analysis was used to analyze the regional differences in the influence of major influencing factors and to identify universal key factors and local dominant factors.

[0080] The key factors and local dominant factors were submitted to experts for review, and their weights were adjusted or qualitative constraints were added.

[0081] Establish a feedback mechanism to periodically reassess the weights of influencing factors based on new data.

[0082] Key factors and local dominant factors are mapped to the agent, the model is trained to predict the correlation between influencing factors and agent behavior, and the influencing factors are directly associated with the agent;

[0083] If the main factors affecting land use planning involve land selection and investment scale, then these main factors should be associated with the intelligent agents represented by the development companies.

[0084] If the main factors affecting land and space planning involve living environment, public services, and employment opportunities, then these main factors should be associated with intelligent agents representing the public and government and functional departments.

[0085] If the main factors affecting land and space planning involve natural systems (such as water bodies and forests), then these main factors should be associated with the intelligent entities of the government and its functional departments and representatives of the National People's Congress and the Chinese People's Political Consultative Conference.

[0086] Based on the weight values ​​of different influencing factors, the speaking time of the agents representing different influencing factors is adjusted. The larger the weight value, the longer the speaking time of the agent it represents.

[0087] After adjusting the weight values ​​of different agents, a dynamic game algorithm is used to prompt the multiple agents to continuously adjust the initial land space transformation plan in multiple rounds of interactive discussions, thus forming a secondary planning scheme.

[0088] During the multiple rounds of speaking, the speaking data of the various intelligent agents is collected in real time through monitoring equipment.

[0089] The speech data is analyzed and detected using detection models and algorithms;

[0090] Based on the analysis and detection results, dynamic intervention is carried out on the violating AI agent, suspending the AI ​​agent's speaking privileges, removing the speaking data of the violating AI agent regarding the initial spatial planning, and triggering manual review.

[0091] The secondary planning scheme is input into the BIM and GIS simulation platform. A three-dimensional model is constructed through the BIM and GIS simulation platform to determine the planning entity elements, the optical attribute information of each type of planning entity element, and the geometric parameter information.

[0092] The 3D model is converted and optimized in holographic format to obtain the converted holographic model framework;

[0093] Determine the planning entity elements of the holographic model framework, the optical attribute information of each type of planning entity element, and the geometric parameter information;

[0094] Based on the geometric parameters, construct parametric geometric models corresponding to various planning entity elements;

[0095] Obtain the texture map of the parametric geometry model and map the texture map onto the parametric geometry;

[0096] The aforementioned geometric parameters are imported into the holographic model framework to generate an initial holographic planning model.

[0097] Based on the initial holographic planning model, and combined with the optical attribute information of each type of entity element, a shader is used for lighting rendering to generate a target holographic planning model with realistic lighting effects.

[0098] Import the holographic planning model into a predetermined scenario, and perform dynamic simulation and deduction based on the predetermined scenario;

[0099] The simulation results are analyzed from multiple dimensions, including economic, ecological, social maintenance, and policy risk dimensions, and the results are visualized.

[0100] Based on the visualization results, the multi-party intelligent agents vote on the quadratic programming scheme to determine whether to adopt the quadratic programming scheme.

[0101] If so, adopt a secondary planning scheme and use the secondary planning scheme as the final planning scheme;

[0102] If not, it will drive multiple intelligent agents to continue speaking on the secondary planning scheme in multiple rounds until a final planning scheme is formed.

[0103] Example 2

[0104] Based on the above embodiment 1, the holographic planning model is imported into a predetermined scenario, and dynamic simulation and deduction are performed according to the predetermined scenario.

[0105] refer to Figure 8 As shown, the predetermined scenarios mainly include development-oriented scenarios, constraint-driven scenarios, emergency scenarios, and policy intervention scenarios.

[0106] By observing multi-scenario simulations using 3D and holographic models, planners representing all parties can conduct visual analysis of the simulated scenarios using holographic visualization and virtual reality (VR) devices.

[0107] Furthermore, in some other embodiments, in addition to the combined use of holographic visualization analysis and VR, the following visualization methods are also included:

[0108] I. Visualization of Two-Dimensional Space

[0109] 1. Thematic Maps

[0110] Purpose: To demonstrate spatial distribution differences under specific scenarios (such as land use type, ecological sensitivity).

[0111] Tools: ArcGIS, QGIS, Mapbox.

[0112] Example:

[0113] Heat map: Shows urban expansion hotspots under different development intensities.

[0114] Tiered color map: Comparing the distribution of carbon emission intensity between the "baseline scenario" and the "carbon neutral scenario".

[0115] 2. Dynamic Time-Series Maps

[0116] Application: To simulate spatial changes over time (such as land use evolution in 2030 and 2050).

[0117] Tools: Leaflet Timeline plugin, Google Earth Engine.

[0118] Example: An animation showing the annual expansion of construction land, overlaid with the changing trajectory of ecological red lines.

[0119] II. Statistics and Chart Visualization

[0120] 1. Radar chart and parallel coordinate graph

[0121] Purpose: To compare the trade-offs between different indicators under multiple scenarios (such as economic-ecological-social equity).

[0122] Tools: Python Matplotlib, Tableau.

[0123] Example: The radar chart shows the indicator scores for the scenarios of "development priority", "ecological priority" and "balanced development".

[0124] 2. Box plot and scatter plot

[0125] Purpose: To analyze the statistical distribution and outliers of scenario simulation results.

[0126] Tools: R ggplot2, Excel.

[0127] Example: Box plots compare the dispersion of carbon emissions in different regions under different scenarios.

[0128] III. Storytelling and Narrative Visualization

[0129] 1. Story Maps

[0130] Purpose: To illustrate the evolutionary logic of a planning scenario by combining maps, text, and images.

[0131] Tools: ArcGIS StoryMaps, Mapbox Storytelling.

[0132] Example: Present the complete narrative of "current problems → scenario simulation → planning scheme" in chapters.

[0133] 2. Animation and Video Deduction

[0134] Purpose: To visually represent spatial changes over a long period of time.

[0135] Tools: Blender animation module, Adobe After Effects.

[0136] Example: Create an animation of a city's expansion over a century, highlighting the impact of major policy milestones.

[0137] IV. Digital Twins and Real-Time Visualization

[0138] 1. Digital Twin Cities

[0139] Application: Integrating IoT sensor data to dynamically simulate real-world responses.

[0140] Tools: Bentley iTwin, Siemens MindSphere.

[0141] Example: Real-time display of the impact of traffic flow changes on land use functions, supporting dynamic strategy adjustments.

[0142] Example 3

[0143] Based on the above embodiments, to ensure fairness and compliance in the multi-party intelligent agent game interaction process, a monitoring device is provided during the game process to monitor the fairness and compliance of the multi-party intelligent agents' speech process. The monitoring process includes:

[0144] Acquire voice, video, or text data from multiple intelligent agents during the game process, and preprocess the data. The acquisition tools include:

[0145] Voice: FFmpeg (audio stream capture), WebRTC (real-time communication).

[0146] Text: Kafka Connect (log collection), OCR (scanning files and converting them to text).

[0147] Video: RTSP streaming media server, OpenCV (real-time frame extraction).

[0148] Multi-source data is input into a multimodal analysis layer, which then analyzes the data, including:

[0149] Speech analysis:

[0150] Speech recognition: Whisper (OpenAI) to text, supports multiple languages.

[0151] Sentiment analysis: SpeechBrain detects tone of voice (anger / cooperation tendency) and marks potential conflicts.

[0152] Text analysis:

[0153] Compliance review:

[0154] Rule Engine: Drools parses policy text (such as "development in wetlands is prohibited") and matches it with agent proposals.

[0155] Semantic similarity: Sentence-BERT matches terms with interactive text to calculate the probability of violation.

[0156] Video analysis:

[0157] Behavior recognition:

[0158] Motion detection: YOLOv8 recognizes gestures (such as pointing at an ecological red line map);

[0159] Scenario Understanding: Segment Anything (SAM) divides the area of ​​illegal construction.

[0160] Compliance comparison: Overlay the land parcels in the video with the planning vector map to detect boundary violations. YOLOv8 detects construction boundary violations, and GIS overlays compare the ecological red line.

[0161] The dynamic response layer receives the analysis results from the multimodal analysis layer in real time and performs dynamic interventions based on the analysis results, specifically including:

[0162] Real-time intervention:

[0163] Voice / Text: Automatically send compliance warnings (such as "The current proposal violates Article 5 of the Wetlands Protection Law").

[0164] Video: AI generates AR annotations (such as flashing red boxes marking out-of-bounds areas).

[0165] Offline handling: Reduce the decision-making weight or resource quota of the violating AI agent, remove its speech data, and conduct manual intervention.

[0166] Furthermore, in some embodiments, the multimodal analysis layer establishes multiple compliance parameters for fairness detection, such as the reward Gini coefficient, Shapley bias, and resource Herfindahl index. It adopts the 3σ principle to mark agents whose rewards exceed the mean ± 3 standard deviations, and uses the Isolation Forest to identify agents with abnormal behavior patterns. When detecting compliance, a compliance rule base is established.

[0167] Example 4

[0168] Based on the above embodiment 1, both initial spatial planning and the acquisition of influencing factors for territorial spatial planning involve the acquisition of multi-source data. Multi-source data can be categorized into structured data, unstructured data, and real-time streaming data. In addition to literature reviews and policy documents, the influencing factors can be classified and collected using the following methods:

[0169] 1. Natural Resources and Environment

[0170] Key factors:

[0171] Topography and landforms (slope, elevation, geological hazard risk);

[0172] Water resource distribution (rivers, lakes, groundwater level);

[0173] Ecological sensitivity (biodiversity, soil erosion, carbon sequestration capacity);

[0174] Climate conditions (rainfall, frequency of extreme weather);

[0175] Collection method:

[0176] Remote sensing images (Sentinel, Gaofen satellite) are used to extract vegetation cover and land use change.

[0177] Geological exploration obtains data on underground rock strata and mineral distribution.

[0178] Sensor networks (weather stations, hydrological stations) monitor temperature, humidity, and water quality in real time.

[0179] 2. Economy and Industry

[0180] Key factors:

[0181] Industrial structure (proportion of agriculture, industry, and service industries).

[0182] Transportation network (road density, logistics efficiency).

[0183] Economic value of land (GDP per unit area, land transfer price).

[0184] Energy supply and demand (electricity, oil and gas pipeline layout).

[0185] Collection method:

[0186] Economic Statistics Yearbook: Obtain GDP, population, and industry data.

[0187] POI big data (Gaode, Baidu Maps) analysis of the distribution of commercial and industrial facilities.

[0188] Enterprise research: Gathering planning intentions for industrial parks and major projects.

[0189] 3. Society and Population

[0190] Key factors:

[0191] Population distribution (density, age structure, migration trends).

[0192] Public service facilities (school and hospital coverage).

[0193] Cultural heritage (historical districts, traditional villages).

[0194] Social equity (accessibility of residential areas for low-income groups).

[0195] Collection method:

[0196] Mobile signaling data: tracking population movement heatmaps.

[0197] Social media sentiment analysis (Weibo, WeChat) identifies public demands.

[0198] Community questionnaire survey: to collect residents' satisfaction with the facilities.

[0199] Example 5

[0200] As a preferred approach, in the process of prompting multiple intelligent agents to continuously adjust the initial land spatial planning and form a secondary planning scheme through multiple rounds of interactive discussions, an interest demand model and a decision-making strategy model are introduced. Specifically, the application process of the interest demand model includes:

[0201] Role identification: List all stakeholders (government and functional departments, technical expert teams, public representatives, development companies, etc.).

[0202] Request extraction: Extract core interests and objectives through interviews and text analysis (NLP).

[0203] Model selection: Match interest-based models based on conflict type (competition / cooperation) and data availability.

[0204] Parameter calibration: Assigning values ​​based on historical data, policy documents, or expert knowledge.

[0205] Simulation and deduction: Outputs heatmaps of interest satisfaction and conflict under different simulated scenarios.

[0206] Strategy optimization: Adjust the planning scheme (such as modifying the plot ratio) to Pareto optimality.

[0207] For example, the interest demand model of local government agents takes into account factors such as fiscal budget constraints, goals for improving people's livelihoods, and factors that enhance the city's image.

[0208] The decision-making strategy model is based on past similar land and space planning cases and policy guidance to formulate strategies such as fund allocation and project priority ranking. The public representative intelligent agent constructs a preference model based on big data analysis of resident feedback, focusing on the degree of resolution of issues closely related to daily life, such as water supply, property management, and the environment.

[0209] Furthermore, based on the Pearson correlation coefficient, stakeholders or non-stakeholders in the game process are identified, such as representatives of the public and development companies. Based on the correlation of interests, different game learning algorithms are dynamically switched during the game process, specifically including:

[0210] Cooperative game theory: Agents maximize common benefits through alliances, using value decomposition network algorithms.

[0211] Non-cooperative game theory: individual rational decision-making (Nash equilibrium), such as the land use game between enterprises and the government, using the Nash Q learning algorithm.

[0212] Hybrid motivation game: cooperation and competition coexist. The upper level selects the cooperation / competition mode, and the lower level calls the corresponding algorithm, such as QMIX when cooperating and MADDPG when competing.

[0213] Example 6

[0214] Before constructing the initial national spatial plan, it is necessary to determine the national spatial boundaries, specifically:

[0215] I. Core Types of Planning Boundaries

[0216] Ecological protection red line:

[0217] Protect key ecological functional areas such as biodiversity conservation, water source conservation, and soil and water conservation.

[0218] Permanent basic farmland:

[0219] To ensure food security, high-quality, contiguous arable land should be designated.

[0220] Urban development boundaries:

[0221] Control the disorderly expansion of cities and make intensive and efficient use of construction land.

[0222] Other specific boundaries:

[0223] Such as historical and cultural protection zones, mineral development zones, and disaster risk zones.

[0224] II. Methods for Determining Planning Boundaries

[0225] 1. Data-driven analysis

[0226] Multi-source data integration:

[0227] Basic geographic data (remote sensing imagery, topography, soil, climate).

[0228] Socioeconomic data (population, GDP, industry distribution).

[0229] Ecological data (biodiversity, water source protection areas, ecological sensitivity assessment).

[0230] Spatial analysis techniques:

[0231] Use GIS for overlay analysis, buffer analysis, and suitability evaluation.

[0232] 3. Multi-objective collaborative decision-making

[0233] Game Theory and Trade-offs:

[0234] Balancing the conflicts between ecological protection, agricultural development, and urban construction.

[0235] The optimal solution set is generated through Pareto front analysis (multi-objective optimization).

[0236] Public participation and expert review:

[0237] Collect stakeholder opinions through hearings and questionnaires.

[0238] Expert consultation ensures the technical rationality and policy compliance.

[0239] Furthermore, before conducting multi-agent game theory, it is necessary to train the multi-agent model, specifically including:

[0240] I. Training Core Technology Framework

[0241] 1. Environmental Modeling

[0242] Spatial Map Building

[0243] GIS data is used to generate gridded or vectorized environments, including layers of terrain, resource distribution (farmland, water sources), and infrastructure.

[0244] Example: 1000×1000 grid, each grid cell stores attributes (land use type, ecological sensitivity, economic value).

[0245] Dynamic rule engine

[0246] Coding policy constraints (such as "development is prohibited within ecological red lines") and physical rules (pollution diffusion models).

[0247] Example: Identifying the scope of ecological protection red lines using an ecological sensitivity model.

[0248] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A multi-scenario simulation and deduction method based on territorial spatial planning, characterized in that, include: Collect land and space information data and initial land and space plans, and construct an intelligent agent representing multiple planning personnel based on the initial spatial plans; An attention-based speaking rule is adopted to drive multiple agents to speak in multiple rounds regarding the initial land use plan; Through dynamic game theory algorithms, multiple intelligent agents are prompted to continuously adjust the initial land space transformation plan through multiple rounds of interactive discussions, thus forming a secondary planning scheme. A holographic model is constructed based on a quadratic programming scheme, and multi-scenario simulations are performed based on the holographic model.

2. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 1, characterized in that, Before the process of driving multiple intelligent agents to speak in multiple rounds regarding the initial land use plan, it also includes: By obtaining historical interaction data of multiple agents from the database, the Pearson correlation coefficients among the multiple agents are calculated. The correlation between agents is determined based on the value of the Pearson correlation coefficient; Different dynamic game algorithms are used based on the relationships between agents.

3. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 1, characterized in that, The process of constructing a holographic model based on a quadratic programming scheme, followed by multi-scenario simulation and deduction based on the holographic model, further includes: The agent obtains the simulation results and determines whether to adopt a secondary planning scheme based on the simulation results. If so, adopt a secondary planning scheme and use the secondary planning scheme as the final planning scheme; If not, it will drive multiple intelligent agents to continue speaking on the secondary planning scheme in multiple rounds until a final planning scheme is formed.

4. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 1, characterized in that, The construction of the holographic model based on the quadratic programming scheme, and the multi-scenario simulation based on the holographic model, include: The secondary planning scheme is input into the BIM and GIS simulation platform, a three-dimensional model is constructed through the BIM and GIS simulation platform, and a holographic planning model is constructed based on the three-dimensional model. Import the holographic planning model into a predetermined scenario, and perform dynamic simulation and deduction based on the predetermined scenario; The simulation results are analyzed from multiple dimensions, and the results are then visualized.

5. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 1, characterized in that, Before adopting attention-based speaking rules to drive multiple rounds of speaking among the multi-party agents regarding the initial land use plan, the following also applies: Obtain the influencing factors of land and space planning, and map these influencing factors to multi-party intelligent agents; Based on the weights of the influencing factors, the speaking weights of the multi-party agents are assigned, and different speaking times are assigned to the multi-party agents based on their speaking weights.

6. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 5, characterized in that, Before obtaining the influencing factors of territorial spatial planning, the following are also included: Based on a literature review and policy documents, a pool of potential influencing factors was compiled. The contribution of each factor in the influencing factor pool is calculated using GA and machine learning models. Based on the contribution values ​​of each factor, list several major influencing factors; GWR analysis was used to analyze the regional differences in the influence of major influencing factors and to identify universal key factors and local dominant factors. The key factors and local dominant factors were submitted to experts for review, and their weights were adjusted or qualitative constraints were added. Establish a feedback mechanism to periodically reassess the weights of influencing factors based on new data.

7. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 3, characterized in that, The construction of the holographic planning model based on the three-dimensional model includes: The 3D model is converted and optimized in holographic format to obtain the converted holographic model framework; Determine the planning entity elements of the holographic model framework, the optical attribute information of each type of planning entity element, and the geometric parameter information; Based on the geometric parameters, construct parametric geometric models corresponding to various planning entity elements; Obtain the texture map of the parametric geometry model and map the texture map onto the parametric geometry; The aforementioned geometric parameters are imported into the holographic model framework to generate an initial holographic planning model. Based on the initial holographic planning model, and combined with the optical attribute information of each type of entity element, a shader is used for lighting rendering to generate a holographic planning model with realistic lighting effects.

8. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 1, characterized in that, The generation of the initial spatial plan includes: Integrate multi-source data and preprocess the multi-source data; Obtain preprocessed multi-source data and mine spatial patterns from the multi-source data using association rule algorithms; By using a convolutional neural network to combine historical successful planning schemes in the database with spatial patterns from multiple sources, an initial spatial plan that conforms to the specifications is generated.

9. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 1, characterized in that, The aforementioned attention-based speaking rules, which drive multiple agents to speak in multiple rounds regarding the initial land use plan, include: During multiple rounds of speaking, the speaking data of the various intelligent agents is collected in real time through monitoring equipment; The speech data is analyzed and detected using detection models and algorithms; Based on the analysis and detection results, dynamic intervention is carried out on the non-compliant intelligent agents.

10. The multi-scenario simulation and deduction method based on territorial spatial planning according to claim 3, characterized in that, The simulation results are analyzed from multiple dimensions, including economic, ecological, social maintenance, and policy risk dimensions.

Images