An old town continuous piece reconstruction scheme generation method, device, equipment and medium

Abstract

This invention discloses a method for generating redevelopment plans for contiguous old towns, comprising: acquiring redevelopment-related data of the area to be redeveloped; extracting the current status characteristics and redevelopment orientation type of the area from the redevelopment-related data; wherein the current status characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics, and location characteristics; inputting the current status characteristics into a pre-trained indicator prediction model corresponding to the redevelopment orientation type, and outputting suggested land use proportion values ​​and suggested plot ratio values ​​for each land use type; building a reinforcement learning environment, training an agent in the reinforcement learning environment, and generating redevelopment plans that meet the constraints; wherein the constraints are the suggested land use proportion values ​​and suggested plot ratio values ​​for each land use type. This invention enables efficient and accurate generation of redevelopment plans for areas to be redeveloped, improving the scientific nature and forward-looking nature of urban renewal and redevelopment planning.

Description

Technical Field

[0001] This invention relates to the field of urban redevelopment technology, and in particular to a method, apparatus, equipment and medium for generating integrated redevelopment plans for old towns. Background Technology

[0002] Urban renewal and spatial transformation are important means to optimize the development pattern of national land space, improve land use efficiency, and promote high-quality urban development. Traditional urban renewal plans are usually based on the experience of planners, combined with current survey data, higher-level planning requirements, and manual judgment to determine land use structure, development intensity, and spatial layout. The final result is formed through multiple rounds of plan comparison. With the gradual application of big data, artificial intelligence and other technologies in the field of urban and rural planning, existing technologies have begun to try to use data models to assist in indicator calculation, spatial simulation and plan optimization, which has improved the efficiency and precision of planning to a certain extent and provided digital support for spatial transformation decisions.

[0003] However, existing methods for generating space renovation plans still rely excessively on the professional experience and subjective judgment of planners, lack sufficient exploration of patterns in historical renovation cases, and lack objective data-driven macroeconomic indicators, making it difficult to guarantee the scientific validity and stability of the plans. Summary of the Invention

[0004] This invention provides a method for generating integrated urban renewal plans, which can efficiently and accurately generate renewal plans for areas to be renovated, thereby improving the scientific nature and forward-looking nature of urban renewal planning.

[0005] In a first aspect, embodiments of the present invention provide a method for generating a contiguous urban renewal plan, comprising: Obtain relevant data on the area to be renovated, and extract the current status characteristics and renovation orientation type of the area to be renovated from the relevant data; wherein, the current status characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics and location characteristics; The existing background characteristics are input into the pre-trained indicator prediction model corresponding to the transformation guidance type, and the suggested land use ratio and plot ratio corresponding to each land use type are output. A reinforcement learning environment is established, and an agent is trained in the reinforcement learning environment to generate a transformation scheme that meets the constraints; the constraints are the suggested land use ratio and the suggested plot ratio corresponding to each land use type.

[0006] Furthermore, the extraction of the current baseline characteristics and transformation orientation type of the area to be transformed from the transformation-related data includes: The topic extraction of the transformation-related data is performed using the LDA topic model to obtain the topic probability distribution vector of the area to be transformed. Based on the topic probability distribution vector, current background features are extracted and then standardized. Clustering algorithms are used to perform unsupervised clustering of the standardized background features, and the transformation orientation type is determined based on the clustering results; the transformation orientation types include historical preservation-led, industrial upgrading-led, livable living-led, and comprehensive quality improvement-led.

[0007] Furthermore, the training process of the indicator prediction model includes: Extract feature vectors and target vectors from historical successful redevelopment cases; wherein, the feature vectors include current background characteristics and redevelopment guidance types, and the target vectors include land use type proportions and average plot ratios; For each type of transformation guidance, a corresponding indicator prediction model is constructed, and the indicator prediction model is trained based on the feature vector and the target vector; The training results were validated using root mean square error and coefficient of determination. Once the validation was successful, the trained indicator prediction models corresponding to each type of transformation guidance were obtained.

[0008] Furthermore, the construction of the reinforcement learning environment includes: A grid system is established with the boundaries of the land parcel to be transformed as the scope, dividing the land parcel to be transformed into several grid units of a preset size; A multi-channel feature vector is constructed for each grid cell, and the feature vectors of all grid cells are stacked to form a three-dimensional tensor, which serves as the observation state of the agent. Define a hierarchical action space, where the first action space is used to select the grid cell to be operated on, and the second action space is used to select a combination scheme of land use and development intensity for the grid cell; Design a comprehensive reward function, which consists of reward for indicator compliance, spatial quality, public interest, economic efficiency, and negative reward for violation of constraints, as well as their respective weight combinations.

[0009] Furthermore, the multi-channel feature vector includes one-hot encoding of the current land use type, development intensity features, constraint identifiers, accessibility features, and economic and environmental features.

[0010] Furthermore, the training and learning of the intelligent agent employs the PPO algorithm, including: Construct a policy network and a value network, and set the hyperparameters of the PPO algorithm; wherein the policy network is used to output the probability distribution for each action, and the value network is used to evaluate the value of the current state; The system iteratively executes interactive sampling, experience storage, batch learning, advantage estimation, and policy optimization steps, updates the policy network parameters, and learns the optimal decision policy. During training, training metrics are monitored in real time, and the model with the best training metrics is selected as the optimal decision model.

[0011] Furthermore, the method also includes: The generated renovation plan will be evaluated from multiple dimensions, including the achievement of key indicators, spatial performance, protection of public interests, and economic feasibility. The output evaluation results meet the preset standards for the renovation plan.

[0012] Secondly, embodiments of the present invention provide an apparatus for generating integrated urban renewal plans, comprising: The data extraction module is used to acquire relevant data on the area to be renovated, and extract the current status characteristics and renovation orientation type of the area to be renovated from the relevant data; wherein, the current status characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics and location characteristics; The indicator prediction module is used to input the current background characteristics into the pre-trained indicator prediction model corresponding to the transformation guidance type, and output the suggested land use ratio and plot ratio corresponding to each land use type. The scheme generation module is used to build a reinforcement learning environment, train the agent in the reinforcement learning environment, and generate a transformation scheme that meets the constraints; the constraints are the suggested land use ratio and the suggested plot ratio corresponding to each land use type.

[0013] Thirdly, embodiments of the present invention provide an electronic device, comprising: Memory, used to store computer programs; A processor for executing the computer program; Wherein, when the processor executes the computer program, it implements the method for generating integrated renovation schemes for old towns as described in any of the first aspects above.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed, implements the method for generating integrated urban renewal schemes as described in any of the first aspects above.

[0015] Compared with existing technologies, the method for generating integrated urban renewal plans provided by this invention has the following advantages: It acquires relevant data on the area to be renovated, and extracts the existing baseline characteristics and renovation orientation types from the data. The existing baseline characteristics include spatial morphological features, socio-economic features, constraint features, and location features. These features are then input into a pre-trained indicator prediction model corresponding to the renovation orientation type, outputting suggested land use proportions and plot ratios for each land use type. A reinforcement learning environment is established, and an agent is trained within this environment to generate renovation plans that meet the constraints. These constraints are the suggested land use proportions and plot ratios for each land use type. This invention enables efficient and accurate generation of renovation plans for areas to be renovated, enhancing the scientific rigor and forward-looking nature of urban renewal planning. Attached Figure Description

[0016] To more clearly illustrate the technical features of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a method for generating a contiguous renovation plan for old towns, as provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a device for generating integrated renovation schemes for old towns, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to limit the invention.

[0021] Firstly, embodiments of the present invention provide a method for generating integrated urban renewal schemes, see [link to relevant documentation]. Figure 1 This is a flowchart illustrating an embodiment of a method for generating integrated urban renewal schemes provided by the present invention.

[0022] like Figure 1 As shown, the method includes the following steps: S1: Obtain relevant data on the area to be renovated, and extract the current status characteristics and renovation orientation type of the area to be renovated from the relevant data; wherein, the current status characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics and location characteristics; The data related to the redevelopment covers multiple dimensions, including spatial, social, economic, policy, and public demands. Specifically, it includes three aspects: First, representative cases of contiguous urban redevelopment that have been implemented nationwide, achieved excellent results, and have complete data, forming a historical sample database of successful cases to provide data support and experience for subsequent model training and indicator prediction. Second, the current status data of the sites to be redeveloped, including land use status data, building age and quality data, population density and structure data, public service facilities data, transportation accessibility data, environmental quality data, property rights information data, and economic vitality indicators, comprehensively depicting the objective background conditions of the areas to be redeveloped. Third, planning control and development guidance data related to the sites to be redeveloped, including textual data such as higher-level planning documents, relevant policy documents, government work reports, public opinion texts, and resident interview records, from which regional functional positioning, development goals, public expectations, and core policy orientations are extracted, providing a basis for identifying the redevelopment orientation type.

[0023] After completing the collection of multi-source data, the current background characteristics of the area to be transformed are extracted from the relevant data. The current background characteristics are extracted into quantitative indicators from four dimensions: spatial morphology, socio-economic, constraints, and location. Specifically, they include: spatial morphological characteristics, including the land use mix index, average number of building floors, road network density, and green space fragmentation index of the plot before transformation; socio-economic characteristics, including population density, benchmark land price level, and current employment ratio of the tertiary industry; constraints characteristics, including the area of ​​the historical and cultural protection zone and the proportion of preserved buildings; and location characteristics, including distance from the city center, distance from the nearest subway station, and per capita GDP of the administrative district.

[0024] By using the above multi-dimensional indicators, a comprehensive and quantitative expression of the current conditions of the areas to be transformed can be achieved, forming standardized feature vectors that can be directly used as model inputs. At the same time, based on the multi-source data fusion analysis, the transformation guidance type corresponding to the areas to be transformed can be determined, providing prior conditions for subsequent classification indicator prediction and differentiated scheme generation.

[0025] S2: Input the current background characteristics into the pre-trained indicator prediction model corresponding to the transformation guidance type, and output the suggested land use ratio and plot ratio corresponding to each land use type; Based on the redevelopment orientation type of the target redevelopment site, the corresponding pre-trained indicator prediction model is invoked. The current background characteristics are input into the model, and the model learns the correlation between "current background characteristics and planning indicators" and outputs in parallel the suggested area ratio of each major land use type (residential land, industrial land, etc.) and the overall plot ratio of the target site under the specific redevelopment orientation, thereby achieving intelligent prediction of core planning indicators.

[0026] To ensure the compliance of the prediction results, the suggested values ​​of land use ratio and plot ratio output by the model are compared and verified with relevant national and local urban and rural planning technical specifications and standards. For example, it is verified whether the predicted green space ratio meets the lower limit requirements stipulated in the "Urban Residential Area Planning and Design Standard". If the prediction results conflict with relevant specifications and standards, the lower limit requirements of the specifications and standards shall prevail, and the prediction results shall be fine-tuned. The reasons for the adjustment and the adjustment range shall be recorded in detail to ensure that the prediction indicators are both scientific and reasonable and in line with the actual planning requirements.

[0027] S3: Build a reinforcement learning environment, train the agent in the reinforcement learning environment, and generate a transformation scheme that meets the constraints; the constraints are the suggested land use ratio and the suggested plot ratio corresponding to each land use type.

[0028] A reinforcement learning environment adapted to spatial transformation scenarios is built, allowing the intelligent agent to gradually master spatial layout decision-making strategies that meet constraints and take into account multiple objectives through continuous trial and error and iterative learning in the built reinforcement learning environment. The land use ratio and plot ratio suggestions obtained in step S2 are used as core constraints to generate transformation schemes that meet the constraints.

[0029] In summary, this invention breaks through the limitations of traditional renovation scheme generation, which relies on human experience, is inefficient, and highly subjective. It constructs an intelligent path for the entire process of "data extraction - indicator prediction - intelligent generation," enabling efficient and accurate generation of renovation schemes for areas to be renovated. This enhances the scientific and forward-looking nature of urban renewal and renovation planning, and provides reliable technical support for urban renewal work.

[0030] In one optional implementation, the step of extracting the current baseline characteristics and redevelopment orientation type of the area to be redeveloped from the redevelopment-related data includes: The topic extraction of the transformation-related data is performed using the LDA topic model to obtain the topic probability distribution vector of the area to be transformed. Based on the topic probability distribution vector, current background features are extracted and then standardized. Clustering algorithms are used to perform unsupervised clustering of the standardized background features, and the transformation orientation type is determined based on the clustering results; the transformation orientation types include historical preservation-led, industrial upgrading-led, livable living-led, and comprehensive quality improvement-led.

[0031] Specifically, LDA topic modeling is an unsupervised topic modeling method that can automatically mine hidden topic information in transformation-related data. By setting an appropriate number of topics (usually 5-10 topics depending on the scale and data complexity of the area to be transformed), it performs dimensionality reduction processing on massive, multi-dimensional transformation-related data. Each topic is defined by a set of keywords (e.g., Topic 1: Science and Technology Innovation, R&D, Industrial Park; Topic 2: Livable, Community, Elderly Care). Finally, it outputs a topic probability distribution vector, in which each dimension corresponds to the probability value of a topic. This can quantitatively reflect the feature strength of the area to be transformed in each topic dimension, providing a core basis for the subsequent extraction of the current background features.

[0032] Thematic dimensions that are highly correlated with the current baseline characteristics are selected from the thematic probability distribution vector. The probability values ​​of each thematic dimension are used as the core quantitative indicators of the current baseline characteristics and integrated to form the current baseline characteristic vector of the area to be transformed.

[0033] Because the dimensions of different current background features (such as land use intensity, population density, distance from commercial centers, etc.) differ, directly using them for subsequent processing will lead to an imbalance in feature weights. Therefore, it is necessary to standardize the extracted current background features to eliminate the differences in dimensions and improve the accuracy and reliability of feature extraction.

[0034] Finally, unsupervised clustering is performed on the standardized baseline characteristics using a clustering algorithm. For example, fuzzy C-means clustering can be used, which allows samples to belong to multiple transformation-oriented types with different membership degrees. Its clustering objective function is: ; Where C represents the number of clusters, and in this embodiment, C is set to 4, corresponding to the four transformation guidance types; N represents the number of samples, i.e., the total number of samples from the areas to be transformed that need to be clustered; and m is the fuzzy factor. Let i be the quantized feature vector of the i-th region to be transformed. For the first Each cluster center represents a core feature vector of a certain type of transformation orientation. For the sample Cluster centers The degree of membership.

[0035] During clustering, the objective function is optimized iteratively. This process minimizes the function value, making the membership distribution of each sample more reasonable and the characteristics of each cluster center more stable. After clustering, four transformation guidance types are finally output. At the same time, each area to be transformed (i.e., each sample) is assigned a clear type label (selecting the transformation guidance type with the highest membership as the core guidance label of the area) and its corresponding feature vector. This type label and feature vector will serve as the prior conditions for the next indicator prediction model, providing targeted guidance for the subsequent model to accurately output the land use ratio and plot ratio suggestions, ensuring that the indicator prediction is highly compatible with the transformation needs of the area.

[0036] This embodiment can accurately classify the transformation guidance type, and the output type label and feature vector serve as prior conditions, laying the foundation for subsequent targeted indicator prediction and transformation scheme generation, further improving the scientificity and adaptability of the transformation scheme.

[0037] In one optional implementation, the training process of the indicator prediction model includes: Extract feature vectors and target vectors from historical successful redevelopment cases; wherein, the feature vectors include current background characteristics and redevelopment guidance types, and the target vectors include land use type proportions and average plot ratios; For each type of transformation guidance, a corresponding indicator prediction model is constructed, and the indicator prediction model is trained based on the feature vector and the target vector; The training results were validated using root mean square error and coefficient of determination. Once the validation was successful, the trained indicator prediction models corresponding to each type of transformation guidance were obtained.

[0038] Specifically, we first collected a large number of cases of urban renewal projects, selected high-quality cases with good renovation results, complete data and representativeness, and initially built a historical renovation case database. For each case in the database, we extracted the feature vector (X) and the target vector (Y).

[0039] The feature vector (X) consists of two parts: strategic positioning features (X1) and current background features (X2). The strategic positioning features adopt the form of category embedding vectors to quantitatively represent the transformation orientation type to which the case belongs. For example, if the case is historical preservation-led, its strategic positioning feature vector is [1,0,0,0], and if it is industrial upgrading-led, it is [0,1,0,0], and so on. This realizes the numerical representation of the transformation orientation type, which is convenient for model recognition.

[0040] Understandably, in order to eliminate the dimensional differences between features of different dimensions and avoid model training bias and slow convergence due to excessive differences in feature numerical ranges, the extracted feature vectors are Z-Score standardized to map all feature indices to an interval with a mean of 0 and a standard deviation of 1, ensuring that each feature has equal weight and influence in model training.

[0041] The target vector (Y) for each case is the core planning indicator after the case is transformed, which specifically includes the area proportion and average plot ratio of each major land use type. The proportion of major land use types is denoted as [R%, I%, B%, A%, G%] (corresponding to the proportion of residential land, industrial land, commercial service facilities land, public management and public service land, and green space and square land, respectively), and the average plot ratio is denoted as FAR.

[0042] Finally, the feature vectors (X) and target vectors (Y) of all cases are organized into a structured table, as shown in Table 1 below. The cases are then grouped and stored according to their respective transformation orientation types. This facilitates the subsequent construction and training of indicator prediction models for each transformation orientation type, ensuring the model's compatibility with the transformation orientation.

[0043] Table 1. Case Database Structure Decomposition Table

[0044] Furthermore, due to the significant differences in the correlation between the existing background characteristics and planning indicators of different renovation cases, for example, historical preservation-oriented cases pay more attention to the proportion of green space and the preservation of buildings, and their indicator correlation patterns are different from those of industrial upgrading-oriented cases. Therefore, based on the constructed case knowledge base, the cases are grouped according to the renovation orientation type, and corresponding indicator prediction models are constructed for each renovation orientation type.

[0045] First, the case data corresponding to each type of transformation guidance is divided into a training set and a test set according to a preset ratio (e.g., 75%:25%). The training set is used for training and optimizing model parameters, and the test set is used for verifying and evaluating model performance. The core goal of model training is to construct a mapping relationship F(X)→Y, that is, inputting the feature vector (X) of the case and outputting the corresponding target vector (Y), which is the proportion of each major land use type and the average plot ratio.

[0046] During model training, root mean square error (RMSE) and coefficient of determination (R²) are used as evaluation metrics for model training effectiveness. In addition, accuracy thresholds for core indicators are set to ensure that the model prediction performance meets actual planning requirements: the coefficient of determination R² for the main indicators on the test set is >0.7, the root mean square error (RMSE) for land use ratio is <5%, and the root mean square error (RMSE) for plot ratio is <0.5. If the model training results do not reach the above thresholds, the quality of the case data needs to be checked (such as removing cases with missing or abnormal data) until the model training performance meets the standards.

[0047] Once the model training effect reaches the preset standard, training is stopped, and the trained indicator prediction model corresponding to each type of transformation guidance is obtained, which is used for the prediction of core indicators of subsequent target transformation plots.

[0048] The indicator prediction model trained in this embodiment outputs high-precision prediction indicators, providing accurate and reliable constraints for the subsequent construction and modification of reinforcement learning environments.

[0049] In one optional implementation, the construction of the reinforcement learning environment includes: A grid system is established with the boundaries of the land parcel to be transformed as the scope, dividing the land parcel to be transformed into several grid units of a preset size; A multi-channel feature vector is constructed for each grid cell, and the feature vectors of all grid cells are stacked to form a three-dimensional tensor, which serves as the observation state of the agent. Define a hierarchical action space, where the first action space is used to select the grid cell to be operated on, and the second action space is used to select a combination scheme of land use and development intensity for the grid cell; Design a comprehensive reward function, which consists of reward for indicator compliance, spatial quality, public interest, economic efficiency, and negative reward for violation of constraints, as well as their respective weight combinations.

[0050] Specifically, all mandatory constraint data related to the plot to be transformed are loaded, including but not limited to the scope of the historical and cultural protection zone, the coordinates of the location of the preserved buildings, the boundary of the geological disaster zone, the scope of the ecological red line, and the scope of infrastructure control. Through spatial coordinate calibration, format conversion and other processing, these data are converted into spatial data that can be accurately aligned with the subsequent grid system, ensuring that the constraint attributes of each grid unit can be accurately identified, and providing a data foundation for the implementation of subsequent constraint penalty mechanisms.

[0051] Using the precise boundaries of the land parcel to be redeveloped as the scope, a regular grid system is established, dividing the land parcel into uniform grid cells of a preset size (e.g., 50m × 50m). Each grid cell is considered an independently decision-making pixel. Complete initial attributes are calculated and stored for each grid cell to ensure that the agent can fully perceive the grid status. The initial attributes specifically include: current land use nature (e.g., residential, industrial, commercial), building attributes (e.g., building density, building height, building age), constraint condition identifiers (used to mark whether the grid is located in a historical protection zone, whether it contains preserved buildings, whether it belongs to an ecological red line or other constrained area), and location characteristics (e.g., the distance from the grid to the nearest subway station, commercial center, urban core area, etc.). At the same time, based on the spatial positional relationship of the grid, an adjacency relationship graph between grids is established to clarify the adjacent grids of each grid (including adjacent grids on the top, bottom, left, right and diagonal). This adjacency relationship graph is used for subsequent calculation of spatial relationship indicators (e.g., land connectivity, functional agglomeration, etc.) to provide support for the quantification of spatial quality rewards.

[0052] A multi-channel feature vector is constructed for each grid cell. The spatial features and attribute information of the grid system are transformed into a digital form that can be recognized and processed by the agent using state representation, thus constructing the agent's observation state.

[0053] Furthermore, the types of actions that an agent can perform on a single grid cell are clearly defined: a typical action consists of two core parts. First, changing the land use nature of the grid (choosing one from the above six land use types). Second, setting the development intensity of the grid (divided into three levels: low, medium, and high, with different levels corresponding to preset floor area ratio ranges, such as low intensity corresponding to a floor area ratio of 1.0-1.5, medium intensity corresponding to 1.6-2.5, and high intensity corresponding to 2.6-3.5, which can be dynamically adjusted according to the transformation guidance type). A hierarchical action space design is adopted, dividing the action into two layers. The first layer of the action space is used to select the grid cell to be operated on. The agent selects one grid from all valid grid cells as the operation object for the current decision. The second layer of the action space is used to select the specific development plan for the grid cell, that is, combining the above land use nature and development intensity, selecting one from all legal combination plans (such as "residential land + medium intensity development", "green space and plaza land + low intensity development", etc.) as the specific transformation decision for the grid.

[0054] Finally, a comprehensive reward function is designed to quantify the planning objective into immediate feedback, guiding the agent to gradually learn the optimal decision that meets the planning requirements. The comprehensive reward function consists of five core dimensions, each corresponding to a planning objective or constraint. The final immediate reward is constructed by weighted summation. The core dimensions are as follows: Indicator compliance reward: Encourage the transformation plan generated by the agent to approach the target value output by the indicator prediction model in terms of core indicators such as overall land use ratio and plot ratio. The smaller the difference between the actual indicator and the target value, the higher the reward value. Spatial quality reward: Encourage intelligent agents to generate good spatial forms that conform to the spatial design requirements of urban planning. Specific quantitative indicators include moderately mixed land use functions, continuous access to public spaces, reasonable road network structure, and coordinated building layout. The spatial quality reward value is obtained by weighted summation of various quantitative indicators. The higher the value, the better the spatial form. Public interest rewards: guide intelligent agents to value public interests. Specific quantitative indicators include the coverage rate of public service facilities, green space area and accessibility, and protection of historical features. Positive rewards will be given if the indicators are met, and the reward value will be appropriately reduced if the targets are not met. Economic efficiency incentives: Encourage intelligent agents to improve land use efficiency. Specific quantitative indicators include land use efficiency, infrastructure costs, and development revenue, guiding intelligent agents to maximize economic benefits under the premise of compliance. Constraints and penalties: The core objective is to prohibit agents from making decisions that violate mandatory constraints, and to impose severe negative rewards (i.e., penalties) on behaviors that violate mandatory constraints (such as occupying ecological red lines or carrying out renovation and development within historical and cultural protected buildings).

[0055] To further enhance the learning effectiveness of the agent and adapt to the learning focus of different training stages, the initial weights of each dimension are dynamically adjusted, and a course-based learning strategy is adopted: In the early stage of training, the focus is on rewarding compliance with indicators, increasing the weight of the corresponding reward to guide the agent to quickly grasp the constraints of the core planning indicators; in the later stage of training, the corresponding weights are reduced, while the weights of spatial quality rewards and public interest rewards are increased, guiding the agent to further optimize spatial form and safeguard public interests on the basis of meeting indicator requirements, and ultimately learn the optimal decision-making strategy that takes into account multiple objectives.

[0056] The reinforcement learning environment built in this embodiment adapts to the actual needs of spatial transformation, providing reliable support for the subsequent training and transformation scheme generation of intelligent agents, and effectively improving the scientific nature, refinement level and adaptability of the transformation scheme.

[0057] In one alternative implementation, the multi-channel feature vector includes a one-hot code of the current land use type, development intensity features, constraint identifiers, accessibility features, and economic and environmental features.

[0058] Specifically, each channel of the multi-channel feature vector corresponds to a core attribute of the grid. Typical channel settings are as follows: Channels 0-5: One-hot codes for the current land use type, corresponding to 6 land use types: residential land (R), industrial land (I), commercial and service facilities land (B), public management and public service land (A), green space and square land (G), and other land (O). For example, if a grid is currently residential land (R), the code for this channel is [1,0,0,0,0,0]. One-hot coding enables numerical representation of land use types, facilitating rapid identification of grid land use attributes by intelligent agents. Channels 6-7: Development intensity characteristics, specifically including two core indicators: building density and building height, are used to quantify the current development level of the grid and provide a reference for agents to adjust the development intensity.

[0059] Channels 8-10: Constraint identifiers, using binary encoding (0 or 1), respectively marking whether the grid is within a historic protection zone or whether it is the area where a preserved building is located; Channels 11-12: Accessibility features, specifically the distance from the grid to key facilities, obtained after normalization. The smaller the value, the better the accessibility. It is used to quantify the locational advantages of the grid and provide support for the land use layout decisions of the agent. Channels 13-15: Economic and environmental characteristics, including the land value, development potential, and neighborhood environmental characteristics corresponding to the grid, all using standardized quantitative values.

[0060] After constructing the multi-channel feature vector of a single grid, the feature vectors of all grid units are stacked according to their spatial positions to form a three-dimensional tensor (rows × columns × number of channels). The "rows" and "columns" correspond to the spatial coordinates of the grid system, and the "number of channels" corresponds to the 16 feature channels mentioned above. This three-dimensional tensor serves as the complete observation state of the agent and can comprehensively and accurately reflect the overall spatial state, grid attributes, and constraints of the plot to be transformed, ensuring that the agent can make decisions based on complete environmental information.

[0061] In one optional implementation, the training and learning of the agent employs the PPO algorithm, including: Construct a policy network and a value network, and set the hyperparameters of the PPO algorithm; wherein the policy network is used to output the probability distribution for each action, and the value network is used to evaluate the value of the current state; The system iteratively executes interactive sampling, experience storage, batch learning, advantage estimation, and policy optimization steps, updates the policy network parameters, and learns the optimal decision policy. During training, training metrics are monitored in real time, and the model with the best training metrics is selected as the optimal decision model.

[0062] Specifically, firstly, a policy network and a value network are constructed. The core function of the policy network is to take the state tensor of the reinforcement learning environment built above (i.e., a three-dimensional tensor formed by stacking all grid multi-channel feature vectors) as input and output the probability distribution of each action, providing a basis for the agent's action selection. The core function of the value network is to evaluate the value of the current environmental state, that is, to predict the long-term cumulative reward that the agent can obtain in the current state, providing a reference for policy optimization. Secondly, the hyperparameters of the PPO algorithm are set to complete the initialization of the PPO algorithm agent. The hyperparameters include learning rate, discount factor, GAE parameters, and pruning range.

[0063] In this embodiment, the policy network can adopt a convolutional neural network (CNN) structure, which utilizes convolutional layers to effectively extract spatial correlation features and grid attribute correlation features from the state tensor, ensuring that action decisions can fully combine the integrity and correlation of the spatial layout. The number of layers and the size of the convolutional kernels of the CNN can be dynamically adjusted according to the scale of the plot to be transformed and the number of grids, ensuring the accuracy and efficiency of feature extraction. The value network can adopt a multilayer perceptron (MLP) structure, which fits the state features through fully connected layers, adapts to training scenarios of different scales, and ensures the accuracy of state value assessment.

[0064] Furthermore, through iterative cycles of "interactive sampling - experience storage - batch learning - advantage estimation - policy optimization," the agent gradually accumulates optimal decision-making experience and updates network parameters. The specific implementation process is as follows: The first step, interactive sampling: The initialized agent is placed into the constructed reinforcement learning environment for operation. In each training step, the agent first obtains the state tensor of the current environment and inputs it into the policy network. The policy network outputs the probability distribution of all executable actions. Based on this probability distribution, the agent samples an action (i.e., according to the rules of the hierarchical action space, it first selects the grid cell to be operated on, and then selects the combination scheme of land use and development intensity of the grid cell). After the environment executes the action, it updates the attribute state of the corresponding grid cell (such as land use, development intensity, etc.) and calculates the immediate reward corresponding to the action according to the comprehensive reward function, thus completing one interactive sampling.

[0065] The second step is to store the experience: The core information generated by this interaction sampling, including the current state, sampling action, immediate reward, and the new state after the action, is encapsulated into an experience sample and stored in the experience replay buffer. The experience replay buffer adopts a priority replay mechanism, that is, the experience samples are sorted according to the TD error (temporal difference error). The larger the TD error, the higher the value of the decision information contained in the sample, and the higher the probability of it being sampled by subsequent batches of learning. This ensures that the agent learns high-value decision experience first and accelerates training convergence.

[0066] The third step is batch learning: When the number of experience samples accumulated in the experience replay buffer reaches a preset threshold, a batch of experience samples is randomly sampled from it and used as training data for subsequent policy optimization and value network updates. This avoids training fluctuations caused by too few samples in a single sampling and improves the stability of training.

[0067] The fourth step is advantage estimation: using the generalized advantage estimation (GAE) method, based on the sampled empirical samples, the advantage value of each state-action pair is calculated. A positive advantage value indicates that the action is better than the average level, and a negative advantage value indicates that the action is worse than the average level, providing a quantitative basis for subsequent policy optimization.

[0068] The fifth step is policy optimization: The core update formula of the PPO algorithm is adopted to update the parameters of the policy network by maximizing the "pruned alternative target". The core function of the pruning operation is to limit the magnitude of policy updates, keeping the probability ratio of the new policy to the old policy within a preset pruning range, so as to avoid model collapse due to excessive policy update magnitude. At the same time, the parameters of the value network are updated with the state value prediction error as the loss function, so that the value network can more accurately predict the long-term cumulative reward of the current state, providing more reliable support for subsequent action decisions.

[0069] Repeat the above process of "interactive sampling - experience storage - batch learning - advantage estimation - strategy optimization". The number of iterations is usually several thousand to tens of thousands, depending on the training effect. During the training process, the agent gradually eliminates low-reward actions and retains high-reward actions through continuous trial and error. That is, it gradually masters the spatial layout decision-making logic that meets planning indicators, has good spatial quality, and takes into account both public interests and economic efficiency.

[0070] During the training cycle, core training metrics are collected and monitored in real time, including average round reward, achievement rate of key planning metrics, and loss function value. Anomalies in the training process are detected in a timely manner, and hyperparameters or training strategies are adjusted accordingly. An early stopping strategy is implemented: if the evaluation reward no longer significantly improves after several consecutive rounds, training is stopped to avoid ineffective training. Finally, model checkpoints are saved periodically, and the saved model checkpoints are evaluated in a set of independent validation environments. Evaluation metrics in the validation environments are calculated, and the model with the best historical evaluation performance is retained as the optimal decision model for generating subsequent modification schemes.

[0071] Furthermore, the pre-trained optimal decision-making model is loaded, and multiple high-quality, diversified transformation schemes are generated through various differentiated deployment strategies. To achieve scheme diversification, three feasible deployment methods are designed by adjusting the exploratory nature of the strategies. These methods can be used individually or in combination to generate multiple different transformation schemes, meeting the comparison and selection needs of different planning scenarios. Specific deployment methods include: Deterministic generation method: ensures that each decision is the optimal choice under the current state, and generates an "optimal but possibly singular" modification plan; Randomness generation method: In each decision-making process, the agent does not force the selection of the action with the highest probability, but randomly samples according to the action probability distribution output by the policy network to generate multiple modification schemes with obvious layout differences, but all satisfying the core constraints. The bundle search generation method maintains multiple candidate solution sequences simultaneously, gradually expands and filters the optimal candidates, and finally outputs a high-quality and diverse set of solutions.

[0072] This embodiment enables the agent to learn the optimal decision-making strategy that conforms to planning constraints and takes into account multiple objectives, providing reliable model support for the automatic generation of subsequent transformation schemes, and further improving the scientific nature, refinement level and generation efficiency of transformation schemes.

[0073] In an optional implementation, the method further includes: The generated renovation plan will be evaluated from multiple dimensions, including the achievement of key indicators, spatial performance, protection of public interests, and economic feasibility. The output evaluation results meet the preset standards for the renovation plan.

[0074] Specifically, a comprehensive quantitative evaluation is conducted on the multiple modification schemes generated by the intelligent agent to determine whether the schemes meet the preset modification standards. The specific evaluation content is as follows: Key indicator achievement assessment: assess whether the renovation plan meets the preset core planning indicator requirements. The assessment indicators include, but are not limited to, green space ratio, road network density, public service facilities per thousand people, etc. The final score of key indicator achievement is obtained by weighting and averaging all individual scores. Spatial Performance Assessment: The assessment evaluates the spatial form and user experience of the renovation plan. The assessment consists of two core indicators: First, accessibility to public spaces: Based on a grid system, the straight-line distance from each residential grid to the nearest public space such as a park or plaza is calculated; the proportion of residents within a 500-meter radius of public spaces is statistically analyzed, with a pre-set assessment standard of ≥80%. Second, land use mix index: The Shannon diversity index formula is used to calculate the degree of land use mix. A 3×3 grid is used as a calculation window, and the proportion of the six land use types within each window is calculated and substituted into the Shannon diversity index formula to obtain the land use mix index for each window. The average index of all windows is calculated as the land use mix assessment result for the plan. A higher average value indicates a more reasonable mix of land use functions, resulting in a higher score.

[0075] Public interest protection assessment: The assessment evaluates the degree to which the renovation plan protects the public interest. The assessment consists of two core indicators: First, the coverage rate of public service facilities: different service radius standards are set for different types of public service facilities, and the coverage ratio of residential areas within the service radius is calculated. Second, the evaluation of historical features protection: for renovation plans that are dominated by historical protection or include historical buildings, the focus is on assessing the degree of protection of the historical building itself and the surrounding 50-meter control range.

[0076] Economic feasibility assessment: Assess the economic rationality of the renovation plan to ensure its feasibility, including revenue calculation (land transfer revenue, long-term tax revenue), cost calculation (demolition cost, infrastructure construction cost, public service facility construction cost), and investment payback period calculation.

[0077] Finally, the evaluation scores of the four dimensions are weighted and averaged to obtain the comprehensive evaluation score of each transformation plan. The plan that meets the comprehensive score is selected. If the comprehensive evaluation results of all plans fail to meet the transformation standard, the loop optimization mechanism is triggered to readjust the indicator prediction parameters, optimize the agent training strategy, and generate new transformation plans until a plan meets the preset transformation standard.

[0078] In practice, after completing multi-dimensional assessments and selecting compliant solutions, visualization and interactive adjustments can be carried out. The core information of the compliant transformation solutions can be generated into visualized two-dimensional display data, which intuitively presents the spatial layout, land use distribution, development intensity, and various assessment indicator results of the solutions. An interactive adjustment function is also set up, allowing planners to manually adjust local grids based on the intelligently generated compliant solutions, and provide real-time feedback on the comprehensive assessment scores and changes after the adjustments. This helps planners to optimize the solutions and ultimately form the optimal transformation solution that balances intelligent scientific principles with human professionalism.

[0079] This embodiment comprehensively covers the compliance, spatial quality, public interest, and economic feasibility of the plan through quantitative evaluation in four dimensions, ensuring that the generated renovation plan meets planning standards and is ready for implementation, thereby further improving the adaptability and feasibility of the renovation plan.

[0080] Secondly, embodiments of the present invention provide a device for generating integrated urban renewal schemes, see below. Figure 2 This is a schematic diagram of an embodiment of the device for generating integrated renovation schemes for old towns provided by the present invention.

[0081] like Figure 2 As shown, the device includes: The data extraction module 21 is used to acquire relevant data on the area to be transformed, and extract the current background characteristics and transformation orientation type of the area to be transformed from the relevant data; wherein, the current background characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics and location characteristics; The indicator prediction module 22 is used to input the current background characteristics into the pre-trained indicator prediction model corresponding to the transformation guidance type, and output the suggested land use ratio and plot ratio corresponding to each land use type. The scheme generation module 23 is used to build a reinforcement learning environment, train the agent in the reinforcement learning environment, and generate a transformation scheme that meets the constraints; the constraints are the suggested land use ratio and the suggested plot ratio corresponding to each land use type.

[0082] In one optional implementation, the step of extracting the current baseline characteristics and redevelopment orientation type of the area to be redeveloped from the redevelopment-related data includes: The topic extraction of the transformation-related data is performed using the LDA topic model to obtain the topic probability distribution vector of the area to be transformed. Based on the topic probability distribution vector, current background features are extracted and then standardized. Clustering algorithms are used to perform unsupervised clustering of the standardized background features, and the transformation orientation type is determined based on the clustering results; the transformation orientation types include historical preservation-led, industrial upgrading-led, livable living-led, and comprehensive quality improvement-led.

[0083] In one optional implementation, the training process of the indicator prediction model includes: Extract feature vectors and target vectors from historical successful redevelopment cases; wherein, the feature vectors include current background characteristics and redevelopment guidance types, and the target vectors include land use type proportions and average plot ratios; For each type of transformation guidance, a corresponding indicator prediction model is constructed, and the indicator prediction model is trained based on the feature vector and the target vector; The training results were validated using root mean square error and coefficient of determination. Once the validation was successful, the trained indicator prediction models corresponding to each type of transformation guidance were obtained.

[0084] In one optional implementation, the construction of the reinforcement learning environment includes: A grid system is established with the boundaries of the land parcel to be transformed as the scope, dividing the land parcel to be transformed into several grid units of a preset size; A multi-channel feature vector is constructed for each grid cell, and the feature vectors of all grid cells are stacked to form a three-dimensional tensor, which serves as the observation state of the agent. Define a hierarchical action space, where the first action space is used to select the grid cell to be operated on, and the second action space is used to select a combination scheme of land use and development intensity for the grid cell; Design a comprehensive reward function, which consists of reward for indicator compliance, spatial quality, public interest, economic efficiency, and negative reward for violation of constraints, as well as their respective weight combinations.

[0085] In one alternative implementation, the multi-channel feature vector includes a one-hot code of the current land use type, development intensity features, constraint identifiers, accessibility features, and economic and environmental features.

[0086] In one optional implementation, the training and learning of the agent employs the PPO algorithm, including: Construct a policy network and a value network, and set the hyperparameters of the PPO algorithm; wherein the policy network is used to output the probability distribution for each action, and the value network is used to evaluate the value of the current state; The system iteratively executes interactive sampling, experience storage, batch learning, advantage estimation, and policy optimization steps, updates the policy network parameters, and learns the optimal decision policy. During training, training metrics are monitored in real time, and the model with the best training metrics is selected as the optimal decision model.

[0087] In an optional implementation, the apparatus further includes a scheme evaluation module for: The generated renovation plan will be evaluated from multiple dimensions, including the achievement of key indicators, spatial performance, protection of public interests, and economic feasibility. The output evaluation results meet the preset standards for the renovation plan.

[0088] It should be noted that the old town renovation scheme generation device provided in this embodiment of the invention is used to execute all the process steps of the old town renovation scheme generation method in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0089] Thirdly, embodiments of the present invention provide an electronic device, see [link to previous document]. Figure 3 The diagram shown is a structural schematic of an electronic device provided in an embodiment of the present invention.

[0090] like Figure 3 As shown, the device includes: Memory 31 is used to store computer programs; Processor 32 is used to execute the computer program; When the processor 32 executes the computer program, it implements the method for generating integrated renovation schemes for old towns as described in any of the above embodiments.

[0091] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 32 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0092] The processor 32 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0093] The memory 31 can be used to store the computer programs and / or modules. The processor 32 implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory 31 and calling the data stored in the memory 31. The memory 31 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 31 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0094] It should be noted that the aforementioned electronic devices include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 3 The structural diagram is merely an example of the electronic device described above and does not constitute a limitation on the electronic device. It may include more components than shown in the diagram, or combine certain components, or use different components.

[0095] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed, implements the method for generating integrated urban renewal schemes as described in any of the above embodiments.

[0096] It should be understood that the present invention can implement all or part of the processes in the above-mentioned method for generating integrated urban redevelopment schemes, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-mentioned method for generating integrated urban redevelopment schemes. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0097] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. It should be noted that, for those skilled in the art, several equivalent obvious modifications and / or equivalent substitutions can be made without departing from the technical principles of the present invention, and these obvious modifications and / or equivalent substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for generating a comprehensive urban renewal plan, characterized in that, include: Obtain relevant data on the area to be renovated, and extract the current status characteristics and renovation orientation type of the area to be renovated from the relevant data; wherein, the current status characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics and location characteristics; The existing background characteristics are input into the pre-trained indicator prediction model corresponding to the transformation guidance type, and the suggested land use ratio and plot ratio corresponding to each land use type are output. A reinforcement learning environment is established, and an agent is trained in the reinforcement learning environment to generate a transformation scheme that meets the constraints; the constraints are the suggested land use ratio and the suggested plot ratio corresponding to each land use type.

2. The method for generating a contiguous urban renewal plan as described in claim 1, characterized in that, The extraction of the current baseline characteristics and redevelopment orientation type of the area to be redeveloped from the relevant redevelopment data includes: The topic extraction of the transformation-related data is performed using the LDA topic model to obtain the topic probability distribution vector of the area to be transformed. Based on the topic probability distribution vector, current background features are extracted and then standardized. Clustering algorithms are used to perform unsupervised clustering of the standardized background features, and the transformation orientation type is determined based on the clustering results; the transformation orientation types include historical preservation-led, industrial upgrading-led, livable living-led, and comprehensive quality improvement-led.

3. The method for generating a contiguous urban renewal plan as described in claim 1, characterized in that, The training process of the indicator prediction model includes: Extract feature vectors and target vectors from historical successful redevelopment cases; wherein, the feature vectors include current background characteristics and redevelopment guidance types, and the target vectors include land use type proportions and average plot ratios; For each type of transformation guidance, a corresponding indicator prediction model is constructed, and the indicator prediction model is trained based on the feature vector and the target vector; The training results were validated using root mean square error and coefficient of determination. Once the validation was successful, the trained indicator prediction models corresponding to each type of transformation guidance were obtained.

4. The method for generating a contiguous urban renewal plan as described in claim 1, characterized in that, The construction of the reinforcement learning environment includes: A grid system is established with the boundaries of the land parcel to be transformed as the scope, dividing the land parcel to be transformed into several grid units of a preset size; A multi-channel feature vector is constructed for each grid cell, and the feature vectors of all grid cells are stacked to form a three-dimensional tensor, which serves as the observation state of the agent. Define a hierarchical action space, where the first action space is used to select the grid cell to be operated on, and the second action space is used to select a combination scheme of land use and development intensity for the grid cell; Design a comprehensive reward function, which consists of reward for indicator compliance, spatial quality, public interest, economic efficiency, and negative reward for violation of constraints, as well as their respective weight combinations.

5. The method for generating a contiguous urban renewal plan as described in claim 4, characterized in that, The multi-channel feature vector includes the one-hot code of the current land use type, development intensity features, constraint identifiers, accessibility features, and economic and environmental features.

6. The method for generating a contiguous urban renewal plan as described in claim 1, characterized in that, The training and learning of the agent employs the PPO algorithm, including: Construct a policy network and a value network, and set the hyperparameters of the PPO algorithm; wherein the policy network is used to output the probability distribution for each action, and the value network is used to evaluate the value of the current state; The system iteratively executes interactive sampling, experience storage, batch learning, advantage estimation, and policy optimization steps, updates the policy network parameters, and learns the optimal decision policy. During training, training metrics are monitored in real time, and the model with the best training metrics is selected as the optimal decision model.

7. The method for generating a contiguous urban renewal plan as described in claim 1, characterized in that, The method further includes: The generated renovation plan will be evaluated from multiple dimensions, including the achievement of key indicators, spatial performance, protection of public interests, and economic feasibility. The output evaluation results meet the preset standards for the renovation plan.

8. A device for generating integrated urban renewal plans, characterized in that, include: The data extraction module is used to acquire relevant data on the area to be renovated, and extract the current status characteristics and renovation orientation type of the area to be renovated from the relevant data; wherein, the current status characteristics include spatial morphological characteristics, socio-economic characteristics, constraint characteristics and location characteristics; The indicator prediction module is used to input the current background characteristics into the pre-trained indicator prediction model corresponding to the transformation guidance type, and output the suggested land use ratio and plot ratio corresponding to each land use type. The scheme generation module is used to build a reinforcement learning environment, train the agent in the reinforcement learning environment, and generate a transformation scheme that meets the constraints; the constraints are the suggested land use ratio and the suggested plot ratio corresponding to each land use type.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program; The processor executes the computer program to implement the method for generating integrated urban renewal schemes as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the method for generating integrated urban renewal schemes as described in any one of claims 1 to 7.

Images