A distributed urban-rural planning decision support method and system

Abstract

The application relates to the technical field of space planning simulation, in particular to a distributed urban and rural planning decision support method and system, which comprises the following steps: collecting urban and rural multi-source heterogeneous basic data, mapping the data into distributed local subgraphs, using a graph neural network model to perform feature coding on the local subgraphs to generate space feature vectors, uploading the feature vectors to a central coordination node and calculating the coupling correlation degree between regions, constructing a global multi-objective game matrix according to the correlation degree to perform conflict detection and iterative solving, and generating a final visual planning decision scheme according to the optimal solution after convergence. In the application, a cooperative processing mechanism of distributed graph construction and global game evolution is used, so that the defects that planning lacks overall performance and cannot dynamically respond to the correlation influence between regions caused by data islands in the prior art are solved, and the scientificity and timeliness of urban and rural planning decision are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of spatial planning simulation technology, and in particular to a distributed urban and rural planning decision support method and system. Background Technology

[0002] The field of spatial planning simulation technology involves the use of computer technology to collect, model, and analyze geospatial data to assist in optimizing the layout of urban and rural construction. Traditional distributed urban and rural planning decision support methods involve acquiring basic data using geographic information collection terminals distributed in different locations, transmitting it via network to a single central server, and storing indicators such as land use rate, population density, and infrastructure coverage in the server's database management system. Planners then use this data to perform calculations based on preset linear rules or statistical formulas, and display the resulting static charts or two-dimensional maps on a screen.

[0003] In existing technologies, the data processing logic of each acquisition terminal is independent and lacks horizontal interaction, resulting in massive heterogeneous spatial data being fragmented into isolated information islands. When faced with high-concurrency data inflow, the central server suffers from excessive computational load and is prone to response delays. It ignores the dynamic coupling relationship between planning elements in different areas and cannot predict the chain reaction to the surrounding areas when a single area changes, resulting in the final planning scheme lacking overall coordination and timeliness. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a distributed urban and rural planning decision support method and system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a distributed urban and rural planning decision support method, comprising the following steps:

[0006] S1: Collect topographic data, land ownership data and traffic flow data of urban and rural areas, establish a node list and edge weight matrix based on the geographical adjacency relationship and functional dependency attributes between spatial entities, initialize the node list and construct a distributed local spatial subgraph;

[0007] S2: Input the distributed local spatial subgraph into the graph convolutional neural network model, aggregate node neighborhood information and perform multi-layer convolution and pooling operations to extract deep topological features and implicit association patterns in the local area and generate spatial feature vectors;

[0008] S3: Summarize the spatial feature vectors of multiple edge nodes, calculate the coupling strength between vectors and construct a global coupling correlation matrix, establish a multi-objective game model based on the global coupling correlation matrix, use the Nash equilibrium solution algorithm to perform conflict detection and strategy iteration, and generate globally optimal planning parameters;

[0009] S4: Based on the global optimal planning parameters, perform layer overlay and dynamic reconstruction on the basic geographic base map to map the land layout structure, facility supporting levels and ecological corridor distribution, and generate interactive planning decision schemes.

[0010] As a further aspect of the present invention, step S1 specifically comprises:

[0011] S11: Obtain the topographic data, land ownership data, and traffic flow data; clean and standardize the data; extract the geometric centroid coordinates of the plot units as node identifiers; and assign initial feature values ​​to each node based on the functional attributes of the plots.

[0012] S12: Calculate the geographical Euclidean distance and traffic accessibility index between multiple nodes, filter valid connection edges according to the preset adjacency determination threshold, and calculate the weight value of the valid connection edges in combination with the functional dependency attribute to generate an edge weight adjacency matrix.

[0013] S13: Based on the node list and the edge weight adjacency matrix, the urban and rural areas are divided into multiple interconnected independent units, and a distributed local spatial subgraph is established.

[0014] As a further aspect of the present invention, step S2 specifically includes:

[0015] S21: Obtain the distributed local spatial subgraph, and use the graph convolutional layer of the graph convolutional neural network model to perform weighted aggregation of the features of the central node and its neighboring nodes according to the edge weight adjacency matrix to generate an aggregated feature representation;

[0016] S22: Input the aggregated feature representation into a nonlinear activation function layer to perform high-dimensional mapping and transformation on the node features, and extract the deep semantic information of multiple nodes at different spatial scales through multi-layer superimposed convolution operations;

[0017] S23: Perform a global pooling operation on the node features processed by convolution to compress the dimensionality of the graph structure data while retaining key topological information, and generate spatial feature vectors.

[0018] As a further aspect of the present invention, step S3 specifically comprises:

[0019] S31: Obtain the spatial feature vectors of the edge nodes of the differentiated regions, calculate the cosine similarity and functional complementarity coefficient between any two vectors, quantify the degree of development correlation between differentiated regions, and construct a global coupling correlation matrix;

[0020] S32: Based on the global coupling correlation matrix, define the strategy space of multiple planning subjects, and establish a multi-objective game model with the objective functions of maximizing regional economic benefits and minimizing ecological and environmental costs;

[0021] S33: Substitute the preset planning constraints into the multi-objective game model, and use the distributed Nash equilibrium algorithm to perform multiple rounds of strategy evolution and payoff calculation until the strategy payoff of each subject converges to a stable state, thereby generating the globally optimal planning parameters.

[0022] As a further aspect of the present invention, step S4 specifically comprises:

[0023] S41: Obtain the global optimal planning parameters and the basic geographic base map, and perform attribute changes and geometric corrections on the corresponding map patches in the base map based on the land use type code and development intensity index in the parameters;

[0024] S42: Using the geographic information system rendering engine, generate multi-level thematic layers based on the corrected map patch attributes, and then overlay and merge the thematic layers to generate an interactive planning decision scheme.

[0025] As a further aspect of the present invention, the calculation process of the weight value of the effective connected edge includes:

[0026] Obtain the geographical distance and traffic flow intensity values ​​between nodes, and calculate the node values ​​using the following formula. With nodes Composite weight value between :

[0027] ;

[0028] in, Representative node With nodes The composite weight value between them This represents the distance weighting adjustment factor used to balance geometric proximity and functional connection strength. Representative node Geometric centroid and nodes The geographical Euclidean distance between the geometric centroids, The distance attenuation coefficient represents the range of influence of interactions in the control space. Representative node Corresponding spatial entities and nodes The actual traffic flow values ​​between the corresponding spatial entities This represents the maximum traffic flow within urban and rural areas.

[0029] As a further aspect of the present invention, the aggregation process of the graph convolutional neural network model includes:

[0030] The normalized Laplacian matrix and the node feature matrix of the current layer are obtained. The Chebyshev polynomial is used to approximate the graph convolution kernel. The node feature matrix is ​​then filtered to capture the spatial correlation of nodes in the neighborhood of different orders.

[0031] The filtered features are input into a parameterized modified linear unit to remove redundant negative features and enhance the nonlinear expressive power of the model, generating a high-order topological feature map.

[0032] As a further aspect of the present invention, the construction process of the multi-objective game model includes:

[0033] Obtain the projected economic benefits and ecological damage assessment values ​​for regional development, and construct the planning framework based on the following formula. Comprehensive utility function :

[0034] ;

[0035] in, Representative planning entity In strategy combination The combined utility function value is as follows: Representative planning entity The development strategy adopted Representatives other than the main planning body The strategy combinations adopted by other entities besides themselves Weighting coefficients representing economic returns. Representative planning entity Take strategy Economic return forecast at that time Weighting coefficients representing ecological costs. Representative planning entity Take strategy Ecological loss assessment value at that time, Represents the geographical or functional relationship with the planning entity Adjacent sets of subjects, Representative planning entity With adjacent entities The synergy coefficient between them Representative strategy With strategy The spatial compatibility index values ​​between them.

[0036] As a further aspect of the present invention, the geometric correction process for the patch attributes includes:

[0037] Obtain the plot ratio limit and building density threshold defined in the global optimal planning parameters, stretch or compress the three-dimensional height attribute of the patch according to the limit value, and perform boundary smoothing and topology check on the coverage of the patch according to the threshold.

[0038] If spatial overlap or gaps are detected between adjacent patches, the boundary coordinates are fine-tuned using an interpolation algorithm based on Dirichlet-Rayetsen polygons to generate a seamless layout of planned patches.

[0039] A distributed urban and rural planning decision support system, the system being used to implement the aforementioned distributed urban and rural planning decision support method, the system comprising:

[0040] The distributed spatial subgraph construction module is used to collect topographic data, land ownership data and traffic flow data of urban and rural areas. Based on the geographical adjacency relationship and functional dependency attributes between spatial entities, it establishes a node list and edge weight matrix, initializes the node list, and constructs a distributed local spatial subgraph.

[0041] The deep topological feature extraction module is used to input the distributed local spatial subgraph into the graph convolutional neural network model, aggregate node neighborhood information and perform multi-layer convolution and pooling operations to extract deep topological features and implicit association patterns in the local area and generate spatial feature vectors.

[0042] The multi-objective game optimization module is used to summarize the spatial feature vectors of multiple edge nodes, calculate the coupling strength between vectors and construct a global coupling correlation matrix, establish a multi-objective game model based on the global coupling correlation matrix, use the Nash equilibrium solution algorithm to perform conflict detection and strategy iteration, and generate globally optimal planning parameters.

[0043] The interactive decision-making scheme generation module is used to overlay and dynamically reconstruct the basic geographic base map based on the global optimal planning parameters, map the land layout structure, facility supporting levels and ecological corridor distribution, and generate interactive planning decision-making schemes.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0045] This invention addresses the shortcomings of existing technologies, such as the lack of overall planning due to data silos and the inability to dynamically respond to inter-regional correlations, by introducing distributed spatial graph construction and global game evolution logic. By extracting graph features at the edge and solving conflict games at the center, it achieves deep perception and dynamic collaborative optimization of complex urban and rural spatial systems, effectively improving the scientific nature and timeliness of urban and rural planning decisions and the ability to automatically resolve multi-objective conflicts. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the main process of the distributed urban and rural planning decision support method of the present invention;

[0047] Figure 2 This is a schematic diagram of the distributed local spatial subgraph construction process of the present invention;

[0048] Figure 3 This is a schematic diagram of the spatial feature vector extraction process of the present invention;

[0049] Figure 4 This is a schematic diagram of the global optimal planning parameter generation process of the present invention;

[0050] Figure 5 This is a schematic diagram of the interactive planning and decision-making scheme generation process of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the software-based technical solution is described in detail below with reference to system architecture diagrams and embodiments. It should be understood that the specific embodiments described herein are only for explaining the technical solutions of this invention and do not constitute a limitation on the scope of protection.

[0052] In the description of this invention, the system architecture relationships or data processing flows indicated by terms such as "layer," "module," "interface," "data flow," "client," and "server" are all defined based on the architecture diagram or flowchart corresponding to the embodiments. This way of describing is only used to clearly illustrate the logical relationships between the elements in the technical solution, and not to limit the physical deployment form. The term "multiple" includes two or more technical units, including but not limited to multiple data nodes, processing threads, service instances, or functional components and other scalable elements. The specific number is determined according to the actual business scenario and needs to be specifically specified.

[0053] Please see Figure 1 and Figure 2 This invention provides a technical solution: a distributed urban and rural planning decision support method, comprising the following steps:

[0054] S1: Collect topographic data, land ownership data, and traffic flow data of urban and rural areas. Based on the geographical adjacency relationship and functional dependency attributes between spatial entities, establish a node list and edge weight matrix, initialize the node list, and construct a distributed local spatial subgraph.

[0055] S11: Acquire topographic data, land ownership data, and traffic flow data; clean and standardize the data; extract the geometric centroid coordinates of the plot units as node identifiers; and assign initial feature values ​​to each node based on the functional attributes of the plots.

[0056] S12: Calculate the geographical Euclidean distance and traffic accessibility index between multiple nodes, filter valid connecting edges based on the preset adjacency determination threshold, and calculate the weight value of valid connecting edges in combination with functional dependency attributes to generate an edge weight adjacency matrix.

[0057] The calculation process for the weight values ​​of valid connecting edges includes:

[0058] Obtain the geographical distance and traffic flow intensity values ​​between nodes, and calculate the node values ​​using the following formula. With nodes Composite weight value between :

[0059] ;

[0060] in, Representative node With nodes The composite weight value between them This represents the distance weighting adjustment factor used to balance geometric proximity and functional connection strength. Representative node Geometric centroid and nodes The geographical Euclidean distance between the geometric centroids, The distance attenuation coefficient represents the range of influence of interactions in the control space. Representative node Corresponding spatial entities and nodes The actual traffic flow values ​​between the corresponding spatial entities This represents the maximum traffic flow within urban and rural areas.

[0061] S13: Based on the node list and the edge weight adjacency matrix, divide the urban and rural areas into multiple interconnected independent units and establish a distributed local spatial subgraph.

[0062] First, the scope of the urban and rural area to be planned was defined. Taking a pilot area in a suburban area of ​​eastern China as an example, the coverage area is 45 square kilometers. High-precision digital elevation models (DEMs) with a ground resolution better than 0.5 meters were acquired through low-altitude remote sensing operations using drones equipped with LiDAR (Light Detection and Ranging) as the source of topographic data. Simultaneously, the data was accessed from the land and resources database to extract vector data containing parcel codes, land use properties, and ownership boundaries as land ownership data. Traffic flow data was collected 24 / 7 using induction coils and video surveillance cameras deployed at major road intersections within the area. The above multi-source heterogeneous data was then cleaned and standardized. Specifically, geometric objects with topological errors were removed, and the coordinate system of all spatial data was converted to a projected coordinate system to eliminate the impact of projection distortion on distance calculations. For each independent land parcel unit, its geometric centroid coordinates were calculated. Abstract it as nodes in a graph structure. Based on the functional attributes of the land parcels, digital coding was performed in accordance with the "Guidelines for Land and Sea Use Classification in Territorial Spatial Survey, Planning, and Land Use Control," and combined with the current values ​​of the land parcel area and plot ratio, a dimension was constructed. initial feature vector This completes the initialization and assignment of the node list.

[0063] Subsequently, the effective edge weights are calculated and the edge weight adjacency matrix is ​​constructed. The threshold for the calculation range of geographic Euclidean distance is set to [value missing]. The unit of measurement is 1500 meters, meaning that potential spatial interaction is considered to exist only when the straight-line distance between two nodes is less than 1500 meters. Based on this, the geographical distance and traffic flow intensity between nodes are further obtained, and the following formula is used to calculate the node... With nodes Composite weight value between :

[0064] ;

[0065] in, Representative node With nodes The composite weight value between them; This represents a distance weighting adjustment factor used to balance geometric proximity and functional connection strength; its value typically ranges from [value range missing]. ; Representative node Geometric centroid and nodes The geographical Euclidean distance between the geometric centroids; Distance attenuation coefficient representing the range of influence of interactions in the control space; Representative node Corresponding spatial entities and nodes The actual traffic flow values ​​between the corresponding spatial entities; This represents the maximum traffic flow within urban and rural areas.

[0066] In this process, the distance weight adjustment factor The value is set based on the focus of the regional planning. Considering that the area is in a stage of rapid urbanization, spatial compactness and transportation convenience are equally important. Therefore, it was set through an expert scoring method. Distance attenuation coefficient for the range of influence of space interactions. Based on the characteristics of the Gaussian kernel function, a distance of 500 meters is set, meaning that the contribution of the weight to the calculation will decrease exponentially beyond a distance of 500 meters. (Actual traffic flow data) The maximum traffic flow in urban and rural areas was obtained by statistically analyzing the average traffic flow between functional areas corresponding to the nodes during morning and evening peak hours (7:00-9:00 and 17:00-19:00). It is set as the upper limit of the road's design capacity, namely 2500 standard vehicle equivalents per hour (pcu / h).

[0067] The specific calculation example is as follows: Select residential area nodes within the region. With business district nodes The Euclidean distance between their geometric centroids was calculated. Meters, the actual observed average interactive traffic flow during peak hours pcu / h. Substituting the above parameters into the composite weight value calculation formula, we first calculate the geometric proximity part: Secondly, the calculation function connects the parts: The final composite weight value is obtained. The calculated result of 0.693 indicates that the node... With nodes There is a strong spatial correlation between them, and their coordinated development should be a key consideration in future planning. Based on the calculated weight values, [the following categories are excluded]. Weakly connected edges are used to generate a sparse edge-weight adjacency matrix, and the global network is divided into several local spatial subgraphs, each containing 30 to 50 nodes, to meet the needs of distributed computing. Table 1 shows the initialization characteristics of some nodes and the calculated connection weight data.

[0068] Table 1. Initialization and Connection Weight Data of Nodes in Urban and Rural Areas:

[0069] ;

[0070] As shown in Table 1, N001 and N005 are at a moderate distance and have a large flow, resulting in significant weight values. However, although N012 and N001 are physically connected, they are too far apart and have weak functional connections. As a result, their calculated weight of 0.065 is lower than the preset threshold of 0.1. When constructing a local subgraph, this connection edge will be pruned to ensure the compactness of the subgraph structure and the efficiency of the calculation.

[0071] Please see Figure 1 and Figure 3 S2: Input the distributed local spatial subgraph into the graph convolutional neural network model, aggregate the neighborhood information of nodes and perform multi-layer convolution and pooling operations to extract deep topological features and implicit association patterns in the local region and generate spatial feature vectors.

[0072] S21: Obtain the distributed local spatial subgraph, and use the graph convolutional layer of the graph convolutional neural network model to perform weighted aggregation of the features of the central node and its neighboring nodes based on the edge weight adjacency matrix to generate aggregated feature representation.

[0073] The aggregation process of graph convolutional neural network models includes:

[0074] The normalized Laplacian matrix and the node feature matrix of the current layer are obtained. The Chebyshev polynomial is used to approximate the graph convolution kernel. The node feature matrix is ​​then filtered to capture the spatial correlation of nodes in the neighborhood of different orders.

[0075] The filtered features are input into a parameterized modified linear unit to remove redundant negative features and enhance the nonlinear expressive power of the model, generating a high-order topological feature map.

[0076] S22: The aggregated feature representation is input into a nonlinear activation function layer to perform high-dimensional mapping and transformation on the node features, and deep semantic information of multiple nodes at different spatial scales is extracted through multi-layer superimposed convolution operations.

[0077] S23: Perform a global pooling operation on the node features processed by convolution to compress the dimensionality of the graph structure data while retaining key topological information, and generate spatial feature vectors.

[0078] A Graph Convolutional Neural Network (GCN) is used to perform deep feature mining on the distributed local spatial subgraphs constructed by S1. First, the GCN model architecture is constructed, which includes an input layer, two graph convolutional layers, a non-linear activation layer, and a global pooling layer. The input layer receives the feature matrix of the local spatial subgraph. (which includes) There are nodes, and the feature dimension is... and the normalized Laplace matrix ,in For an adjacency matrix with self-loops, for The degree matrix.

[0079] In the computation of graph convolutional layers, to reduce computational complexity and capture the spatial correlation of nodes within neighborhoods of different orders, Chebyshev polynomials are used to approximate the graph convolution kernel. Specifically, the order of the Chebyshev polynomial is set... This means the model can aggregate information from the 3rd-order neighborhood of the central node. For the... Layer input features Feature extraction is performed using convolution operations, which involve Chebyshev polynomials. and the convolution kernel parameters to be learned The model was trained using the Adam optimizer, with an initial learning rate of 0.005 and a weight decay coefficient of [value missing]. To prevent overfitting, the loss function uses mean squared error (MSE) loss combined with a regularization term to measure the difference between the reconstructed features and the target features.

[0080] The features after convolution filtering are input into a parameterized corrected linear unit (PReLU), i.e. ,in These are learnable parameters (initially set to 0.25). This activation mechanism can remove redundant negative features while retaining some information flow in the negative intervals, thereby enhancing the model's ability to express nonlinear spatial relationships and generating high-order topological feature maps. For example, after the first layer of convolution and activation, the feature dimension increases from the original... Mapped to This study effectively extracted the implicit relationship between land use mixing and transportation accessibility.

[0081] Finally, a global pooling operation is performed on the node features processed by the two convolutional layers. Specifically, a global average pooling strategy is adopted. The feature vectors of each node are averaged along their feature dimensions to generate a spatial feature vector of fixed length (e.g., 128 dimensions). This process compresses complex graph-structured data into compact vector representations while preserving key topological information within local regions. Experimental data show that, when setting the Chebyshev polynomial order... Furthermore, when the output dimension of the convolutional layer is 128, the model achieves an accuracy of 92.4% in identifying regional functions, which is 7.3% higher than the traditional GCN method (85.1% accuracy) that only uses first-order neighborhood aggregation. This indicates that by aggregating multi-order neighborhood information, the model successfully captures deep, long-distance dependencies in urban and rural spaces.

[0082] Please see Figure 1 and Figure 4 S3: Summarize the spatial feature vectors of multiple edge nodes, calculate the coupling strength between vectors and construct a global coupling correlation matrix, establish a multi-objective game model based on the global coupling correlation matrix, use the Nash equilibrium solution algorithm to perform conflict detection and strategy iteration, and generate globally optimal planning parameters.

[0083] S31: Obtain the spatial feature vectors of edge nodes in differentiated regions, calculate the cosine similarity and functional complementarity coefficient between any two vectors, quantify the degree of developmental correlation between differentiated regions, and construct a global coupling correlation matrix.

[0084] S32: Based on the global coupling correlation matrix, the strategy space of multiple planning subjects is defined, and a multi-objective game model is established with the goal of maximizing regional economic benefits and minimizing ecological and environmental costs.

[0085] The process of constructing a multi-objective game model includes:

[0086] Obtain the projected economic benefits and ecological damage assessment values ​​for regional development, and construct the planning framework based on the following formula. Comprehensive utility function :

[0087] ;

[0088] in, Representative planning entity In strategy combination The combined utility function value is as follows: Representative planning entity The development strategy adopted Representatives other than the main planning body The strategy combinations adopted by other entities besides themselves Weighting coefficients representing economic returns. Representative planning entity Take strategy Economic return forecast at that time Weighting coefficients representing ecological costs. Representative planning entity Take strategy Ecological loss assessment value at that time, Represents the geographical or functional relationship with the planning entity Adjacent sets of subjects, Representative planning entity With adjacent entities The synergy coefficient between them Representative strategy With strategy The spatial compatibility index values ​​between them.

[0089] S33: Substitute the preset planning constraints into the multi-objective game model, use the distributed Nash equilibrium algorithm to perform multiple rounds of strategy evolution and payoff calculation until the strategy payoff of each subject converges to a stable state, and generate the globally optimal planning parameters.

[0090] This addresses the common issue of multi-stakeholder conflicts of interest in urban and rural planning. First, it obtains the spatial feature vectors of edge nodes in differentiated regions (such as central urban areas, agricultural protection zones, and ecological conservation areas) output by S2. Then, it uses the cosine similarity formula to calculate the similarity between any two vectors. and The directional consistency between the two regions serves as the basis for the functional complementarity coefficient. If the cosine similarity is close to 1, it indicates a high degree of synergy in the development patterns of the two regions; if it is close to -1, it indicates significant functional mutual exclusion. Based on this, a global coupling correlation matrix is ​​constructed. Matrix elements Quantified the region With the region The degree of developmental correlation between them.

[0091] Subsequently, a multi-objective game model is established. The set of participating entities in the game is defined. These represent planning and management units with different interest orientations. With the objective function of maximizing regional economic benefits and minimizing ecological and environmental costs, the planning entity is constructed based on the following formula. Comprehensive utility function :

[0092] ;

[0093] in, Representative planning entity In strategy combination The numerical value of the comprehensive utility function under the following conditions; Representative planning entity The development strategy adopted; Representatives other than the main planning body The strategy combinations adopted by other entities besides themselves; Weighting coefficients representing economic returns; Representative planning entity Take strategy Economic return forecast at that time; Weighting coefficients representing ecological costs; Representative planning entity Take strategy Ecological loss assessment value at that time; Represents the geographical or functional relationship with the planning entity Adjacent sets of entities; Representative planning entity With adjacent entities The synergy coefficient between them; Representative strategy With strategy The spatial compatibility index values ​​between them.

[0094] In the actual parameter settings of the model, the economic return weight coefficient Set to 0.55, ecological cost weighting coefficient The value is set at 0.45, reflecting the current planning stage's focus on economic development while maintaining ecological balance. (Economic benefit forecast) The ecological loss assessment value is calculated using a GDP output model per unit land area. Calculated using an Ecosystem Service Value (ESV) loss model. Synergy coefficient. Directly taken from the corresponding element in the global coupling correlation matrix. Spatial compatibility index. The range of values ​​is The strategy is determined by identifying land use type conflicts between strategies. A distributed Nash equilibrium algorithm is used for multi-round strategy evolution. In the initial state, each agent randomly selects a strategy, and in each iteration... In the middle, the main body Based on the strategies of other entities Choose the function that maximizes its own utility. Optimal response strategy The specific convergence criterion is: when the rate of change of the strategy parameters of all subjects is less than 0.5% in three consecutive iterations, the system is considered to have reached Nash equilibrium.

[0095] A specific example is as follows: Assume there are adjacent entities A (focusing on industrial development) and B (focusing on ecological living). The initial strategy is: A chooses a plot ratio of 2.5 (high-intensity development), and B chooses a plot ratio of 1.0 (low-density living). The economic benefits for A at this time are... Ecological costs B's economic gains Ecological costs Set the synergy coefficient. Compatibility metrics ;

[0096] Calculate the utility of subject A:

[0097] ;

[0098] Calculate the utility of subject B:

[0099] ;

[0100] In the next round of the game, Player A, in order to increase total utility, adjusts the floor area ratio to 2.0. Reduced to 85 Reduced to 40, compatibility Increased to 0.5;

[0101] Recalculate the utility of A:

[0102] ;

[0103] because Subject A tends to accept the adjusted strategy. Table 2 shows a snapshot of the data during the game iteration process.

[0104] Table 2. Data on the strategy iteration process of the multi-objective game model:

[0105] ;

[0106] As shown in Table 2, after 15 iterations, the total system utility increased from 45.38 to 49.93, and the strategy parameters stabilized at a plot ratio of 1.80 and 1.30. This result indicates that the Nash equilibrium algorithm successfully found the optimal solution that balances economic and ecological considerations.

[0107] Please see Figure 1 and Figure 5 S4: Based on the global optimal planning parameters, the base map is overlaid and dynamically reconstructed to map the land layout structure, facility supporting levels and ecological corridor distribution, and generate interactive planning decision schemes.

[0108] S41: Obtain the global optimal planning parameters and basic geographic base map, and based on the land use type code and development intensity index in the parameters, perform attribute changes and geometric corrections on the corresponding map patches in the base map.

[0109] The geometric correction process for patch attributes includes:

[0110] Obtain the plot ratio limit and building density threshold defined in the global optimal planning parameters, stretch or compress the three-dimensional height attribute of the patch according to the limit value, and perform boundary smoothing and topology check on the coverage of the patch according to the threshold.

[0111] If spatial overlap or gaps are detected between adjacent patches, the boundary coordinates are fine-tuned using an interpolation algorithm based on Dirichlet-Rayetsen polygons to generate a seamless layout of planned patches.

[0112] S42: Using the geographic information system rendering engine, multi-level thematic layers are generated based on the corrected map patch attributes, and the thematic layers are overlaid and merged to generate an interactive planning decision scheme.

[0113] After obtaining the globally optimal planning parameters output by S3, the visualization reconstruction and scheme generation stage begins. First, a high-precision base map is loaded, which includes vector layers of existing buildings, road networks, and water systems. Based on the land use type code (e.g., R2 Class II residential land) and development intensity indicators (e.g., plot ratio 1.8, building density 25%) in the parameters, the attributes of the corresponding patches in the base map are changed.

[0114] For the geometric correction process of map patch attributes, rigorous parametric modeling is performed. The floor area ratio limit value defined in the globally optimal planning parameters is obtained. Assuming a plot of land has an area of ​​10,000 square meters, the allowed total building area is 18,000 square meters. If the current building density threshold is set at 25%, the base area must not exceed 2,500 square meters. The average building height is calculated based on this. Meters (approximately 7 layers). The system automatically stretches the three-dimensional height attribute of this plot of land to 21.6 meters.

[0115] Simultaneously, the Dirichlet-Thyson polygon algorithm is used to handle spatial gaps. If undefined blank gaps are detected between adjacent patches after planning adjustments, the centroids of the patches surrounding the gap are extracted as seed points, and Thiessen polygons are generated to cover the area. The gaps are then automatically assigned to adjacent patches according to spatial proximity, achieving seamless connection. For example, irregular corner plots resulting from road boundary adjustments are automatically reclassified as adjacent protective green spaces using this algorithm, ensuring the topological integrity of the planned patches.

[0116] Finally, using the WebGL geographic information system rendering engine, multi-level thematic layers are generated based on the corrected patch attributes. These include: land use planning maps, development intensity control maps, and ecological corridor analysis maps. These thematic layers are then overlaid and merged, with the transparency parameter set to 70% to preserve the base map texture, generating an interactive planning decision-making scheme. This scheme allows users to click on any plot of land and view real-time attribute pop-ups containing detailed indicators such as "planned floor area ratio," "recommended building height," "projected economic output," and "ecological conservation ratio," achieving the digital and visual delivery of planning results.

[0117] A distributed urban and rural planning decision support system is provided. This system is used to execute the aforementioned distributed urban and rural planning decision support method. The system includes:

[0118] The distributed spatial subgraph construction module is used to collect topographic data, land ownership data and traffic flow data of urban and rural areas. Based on the geographical adjacency relationship and functional dependency attributes between spatial entities, it establishes a node list and edge weight matrix, initializes the node list, and constructs a distributed local spatial subgraph.

[0119] The deep topological feature extraction module is used to input distributed local spatial subgraphs into the graph convolutional neural network model, aggregate node neighborhood information and perform multi-layer convolution and pooling operations to extract deep topological features and implicit association patterns within local regions and generate spatial feature vectors.

[0120] The multi-objective game optimization module is used to summarize the spatial feature vectors of multiple edge nodes, calculate the coupling strength between vectors and construct a global coupling correlation matrix, establish a multi-objective game model based on the global coupling correlation matrix, use the Nash equilibrium solution algorithm to perform conflict detection and strategy iteration, and generate globally optimal planning parameters.

[0121] The interactive decision-making scheme generation module is used to overlay and dynamically reconstruct the basic geographic base map based on the global optimal planning parameters, map the land layout structure, facility supporting levels and ecological corridor distribution, and generate interactive planning decision-making schemes.

[0122] The above embodiments illustrate preferred embodiments of the present invention. Any equivalent adjustments to the technical solution based on software engineering methods are within the scope of protection, including but not limited to: implementing algorithm logic using different programming languages, refactoring functional modules into services, adjusting data interaction protocols, and optimizing resource scheduling strategies. Any implementation scheme derived from reasonable modifications to the data processing flow, service call chain, or system architecture layer without departing from the core technology of the present invention should be considered within the protection scope defined by the technical solution of the present invention.

Claims

1. A distributed urban and rural planning decision support method, characterized in that, Includes the following steps: S1: Collect topographic data, land ownership data and traffic flow data of urban and rural areas, establish a node list and edge weight matrix based on the geographical adjacency relationship and functional dependency attributes between spatial entities, initialize the node list and construct a distributed local spatial subgraph; The specific steps of S1 are as follows: S11: Obtain the topographic data, land ownership data, and traffic flow data; clean and standardize the data; extract the geometric centroid coordinates of the plot units as node identifiers; and assign initial feature values ​​to each node based on the functional attributes of the plots. S12: Calculate the geographical Euclidean distance and traffic accessibility index between multiple nodes, filter valid connection edges according to the preset adjacency determination threshold, and calculate the weight value of the valid connection edges in combination with the functional dependency attribute to generate an edge weight adjacency matrix. S13: Based on the node list and the edge weight adjacency matrix, divide the urban and rural areas into multiple interconnected independent units and establish a distributed local spatial subgraph; The calculation process for the weight values ​​of the effective connected edges includes: Obtain the geographical distance and traffic flow intensity values ​​between nodes, and calculate the node values ​​using the following formula. With nodes Composite weight value between : ； in, Representative node With nodes The composite weight value between them This represents the distance weighting adjustment factor used to balance geometric proximity and functional connection strength. Representative node Geometric centroid and nodes The geographical Euclidean distance between the geometric centroids, The distance attenuation coefficient represents the range of influence of interactions in the control space. Representative node Corresponding spatial entities and nodes The actual traffic flow values ​​between the corresponding spatial entities This represents the maximum traffic flow within urban and rural areas. S2: Input the distributed local spatial subgraph into the graph convolutional neural network model, aggregate node neighborhood information and perform multi-layer convolution and pooling operations to extract deep topological features and implicit association patterns in the local area and generate spatial feature vectors; S3: Summarize the spatial feature vectors of multiple edge nodes, calculate the coupling strength between vectors and construct a global coupling correlation matrix, establish a multi-objective game model based on the global coupling correlation matrix, use the Nash equilibrium solution algorithm to perform conflict detection and strategy iteration, and generate globally optimal planning parameters; The specific steps of S3 are as follows: S31: Obtain the spatial feature vectors of the edge nodes of the differentiated regions, calculate the cosine similarity and functional complementarity coefficient between any two vectors, quantify the degree of development correlation between differentiated regions, and construct a global coupling correlation matrix; S32: Based on the global coupling correlation matrix, define the strategy space of multiple planning subjects, and establish a multi-objective game model with the objective functions of maximizing regional economic benefits and minimizing ecological and environmental costs; S33: Substitute the preset planning constraints into the multi-objective game model, and use the distributed Nash equilibrium algorithm to perform multiple rounds of strategy evolution and payoff calculation until the strategy payoff of each subject converges to a stable state, thereby generating the globally optimal planning parameters. The construction process of the multi-objective game model includes: Obtain the projected economic benefits and ecological damage assessment values ​​for regional development, and construct the planning framework based on the following formula. Comprehensive utility function : ； in, Representative planning entity In strategy combination The numerical value of the comprehensive utility function under the following conditions Representative planning entity The development strategy adopted Representatives excluding the main planning body The strategy combinations adopted by other entities besides themselves Weighting coefficients representing economic returns. Representative planning entity Take strategy Economic return forecast at that time The weighting coefficients representing ecological costs. Representative planning entity Take strategy Ecological loss assessment value at that time, Represents the geographical or functional relationship with the planning entity Adjacent sets of subjects, Representative planning entity With adjacent entities The synergy coefficient between them Representative strategy With strategy Spatial compatibility index values ​​between them; S4: Based on the global optimal planning parameters, perform layer overlay and dynamic reconstruction on the basic geographic base map to map the land layout structure, facility supporting levels and ecological corridor distribution, and generate interactive planning decision schemes.

2. The distributed urban and rural planning decision support method according to claim 1, characterized in that, The specific steps of S2 are as follows: S21: Obtain the distributed local spatial subgraph, and use the graph convolutional layer of the graph convolutional neural network model to perform weighted aggregation of the features of the central node and its neighboring nodes according to the edge weight adjacency matrix to generate an aggregated feature representation; S22: Input the aggregated feature representation into a nonlinear activation function layer to perform high-dimensional mapping and transformation on the node features, and extract the deep semantic information of multiple nodes at different spatial scales through multi-layer superimposed convolution operations; S23: Perform a global pooling operation on the node features processed by convolution to compress the dimensionality of the graph structure data while retaining key topological information, and generate spatial feature vectors.

3. The distributed urban and rural planning decision support method according to claim 2, characterized in that, The specific steps of S4 are as follows: S41: Obtain the global optimal planning parameters and the basic geographic base map, and perform attribute changes and geometric corrections on the corresponding map patches in the base map based on the land use type code and development intensity index in the parameters; S42: Using the geographic information system rendering engine, generate multi-level thematic layers based on the corrected map patch attributes, and then overlay and merge the thematic layers to generate an interactive planning decision scheme.

4. The distributed urban and rural planning decision support method according to claim 2, characterized in that, The aggregation process of the graph convolutional neural network model includes: The normalized Laplacian matrix and the node feature matrix of the current layer are obtained. The Chebyshev polynomial is used to approximate the graph convolution kernel. The node feature matrix is ​​then filtered to capture the spatial correlation of nodes in the neighborhood of different orders. The filtered features are input into a parameterized modified linear unit to remove redundant negative features and enhance the nonlinear expressive power of the model, generating a high-order topological feature map.

5. The distributed urban and rural planning decision support method according to claim 3, characterized in that, The geometric correction process for the properties of the image features includes: Obtain the plot ratio limit and building density threshold defined in the global optimal planning parameters, stretch or compress the three-dimensional height attribute of the patch according to the limit value, and perform boundary smoothing and topology check on the coverage of the patch according to the threshold. If spatial overlap or gaps are detected between adjacent patches, the boundary coordinates are fine-tuned using an interpolation algorithm based on Dirichlet-Rayetsen polygons to generate a seamless layout of planned patches.

6. A distributed urban and rural planning decision support system, characterized in that, The system is used to implement the distributed urban and rural planning decision support method according to any one of claims 1-5, and the system includes: The distributed spatial subgraph construction module is used to collect topographic data, land ownership data and traffic flow data of urban and rural areas. Based on the geographical adjacency relationship and functional dependency attributes between spatial entities, it establishes a node list and edge weight matrix, initializes the node list, and constructs a distributed local spatial subgraph. The deep topological feature extraction module is used to input the distributed local spatial subgraph into the graph convolutional neural network model, aggregate node neighborhood information and perform multi-layer convolution and pooling operations to extract deep topological features and implicit association patterns in the local area and generate spatial feature vectors. The multi-objective game optimization module is used to summarize the spatial feature vectors of multiple edge nodes, calculate the coupling strength between vectors and construct a global coupling correlation matrix, establish a multi-objective game model based on the global coupling correlation matrix, use the Nash equilibrium solution algorithm to perform conflict detection and strategy iteration, and generate globally optimal planning parameters. The interactive decision-making scheme generation module is used to overlay and dynamically reconstruct the basic geographic base map based on the global optimal planning parameters, map the land layout structure, facility supporting levels and ecological corridor distribution, and generate interactive planning decision-making schemes.

Images