Urban village reconstruction planning method and system based on generative model

Abstract

The invention provides an urban village reconstruction planning method and system based on a generative model, and the method comprises the steps: generating urban activity conditions after m years according to the determined urban conditions, urban village conditions and a reconstruction scheme set through a first generative model; according to the city activity condition after m years, utilizing a second generative model to calculate the city internal commuting condition after m years and the city facility utilization condition after m years; according to the transformation scheme set, economic income and expenditure conditions after m years are obtained, and according to the economic income and expenditure conditions after m years, the city internal commuting conditions after m years and the city facility utilization conditions after m years, a heuristic algorithm is used for multiple iterations to obtain multiple adjusted target transformation schemes; and according to the multiple adjusted target transformation schemes and the corresponding total economic income and expenditure conditions in the m years, the total city facility utilization conditions in the m years and the total city internal commuting conditions in the m years, performing analysis to obtain an optimal target transformation scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of urban planning, and in particular to a method and system for planning the reconstruction of urban villages based on a generative model. Background Art

[0002] Urban village renovation refers to the comprehensive renovation of urban villages based on the regional socio-economic development and urban master plan and in accordance with urbanization requirements. Actively and steadily promoting urban village renovation in megacities and super-megacities is an important measure to improve people's livelihood and promote high-quality urban development.

[0003] However, large-scale urban village renovation in megacities and super-megacities often requires overall planning among hundreds or even thousands of urban villages, deciding the order and intensity of renovation of urban villages, and coordinating fund allocation within the city. It also requires considering the fund balance, facility layout, and housing price changes after the implementation of the renovation plan, as well as the impact of specific renovation plans on the overall population distribution, job-housing balance, and transportation demand of the city. This poses a huge challenge to traditional planning and decision-making methods.

[0004] The traditional method of forming renovation plans cannot deduce changes in urban elements such as housing prices, facilities, and population, cannot optimize and compare multiple renovation targets at the same time, and cannot meet the needs of planning decisions. Summary of the invention

[0005] The present invention provides a method and system for planning the reconstruction of urban villages based on a generative model, so as to solve the defects that the methods for planning the reconstruction of urban villages in the prior art cannot simultaneously optimize and compare multiple reconstruction targets and cannot meet the needs of planning decision-making.

[0006] The present invention provides a method for planning the reconstruction of urban villages based on a generative model, comprising: Based on the determined urban conditions g, urban village conditions u, and transformation plan set P, the pre-trained first generative model is used to generate the urban activity conditions g m years later. m ; Among them, the first generative model is trained based on the urban condition training samples, the urban village condition samples and the transformation plan sample set; According to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. mAnd the utilization of urban facilities after m years m ; According to the transformation plan set P, the economic revenue and expenditure situation E after m years is obtained m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations and obtain multiple adjusted target transformation plans; The optimal target transformation plan is obtained by analyzing the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total internal urban commuting situation C within m years.

[0007] According to the urban village reconstruction planning method based on the generative model provided by the present invention, the first generative model includes a first network and a second network; Based on the determined urban conditions g, urban village conditions u, and transformation plan set P, the pre-trained first generative model is used to generate the urban activity conditions g m years later. m , including: According to the determined urban condition g, the noise amount is extracted from the standard Gaussian distribution, and then added to the determined urban condition g to obtain the current random urban condition ; Randomize city status with a predefined maximum number of denoising steps Input into the pre-trained first network for processing to obtain the predicted noise ; Randomize city status , the urban village status u and the transformation plan set P are input into the pre-trained second network to calculate the gradient value of the loss function ; Update the noise according to the gradient value , and update the current random city status according to the updated noise to obtain the updated random city status; Iterative execution inputs the updated random city status into the first network and the second network respectively to obtain the next updated random city status step until the maximum number of denoising steps T is 0, and the city activity status m years later is obtained.

[0008] According to the urban village reconstruction planning method based on the generative model provided by the present invention, the training method of the first network includes: Randomly select a city condition training sample from the training sample data set; Determine the random draw time , and from a constant sequence Extract the corresponding constant ; Randomly extracting a random sample noise from a standard Gaussian distribution of the same size as the urban condition training sample, and adding the random sample noise and the constant to the urban condition training sample to obtain a noisy urban condition training sample; The noised urban condition training samples are compared with the randomly selected time Input into the first network to obtain the predicted noise; A loss value is calculated according to the predicted noise and the random sample noise, and the parameters in the first network are updated according to the predicted loss value by using the stochastic gradient descent method until the parameters in the first network converge.

[0009] According to the urban village reconstruction planning method based on the generative model provided by the present invention, the training method of the second network includes: Randomly select a city condition training sample from the training sample data set ; From randomly selected urban condition training samples The corresponding time point is taken as the starting point, and the time point forward to the beginning of the sample data set is taken as the end point, and a time interval is randomly selected ; According to the extraction time interval , obtain known samples of urban village conditions And the corresponding transformation plan sample set ; The randomly selected urban condition training samples Input into the second network to get the predicted transformation time interval 、The situation of urban villages and transformation plans ; Calculate the predicted renovation time interval The time interval between the extraction and the predicted urban village status Compared with known samples of urban villages and predicted transformation plans The difference between the sample and the known transformation plan is used to obtain the second loss value; By using the stochastic gradient descent method, the parameters of the second network are updated according to the second loss value until the parameters of the second network converge.

[0010] According to the urban village reconstruction planning method based on the generative model provided by the present invention, according to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later mAnd the utilization of urban facilities after m years m , including: Define multiple agents in a city. For any agent in the city , records the set of all places it has visited; For any time , pick a random number from a random distribution ; Calculate the exploration probability of agent i ,in, and are two predetermined hyperparameters, is the number of grid points that agent i has visited; If p≥ The control agent performs a return operation, randomly selects a location from the set of locations it has visited, and updates its position; the probability of visiting each location is proportional to the frequency of its previous visits; like , control agent i to perform exploration operations, randomly select a location from the set of other locations in the city, and update its location; the visit probability of each location is proportional to the number of people at the location, and the visit probability of each location is inversely proportional to the distance between two locations; Repeat the above steps for all agents in the city until time Reach the specified time step ; Count the number of visits to the city facilities of interest and calculate the utilization of the city facilities m years later

[0011] Count the number of passes between grids within the city and calculate the commuting situation within the city m years later .

[0012] According to the urban village reconstruction planning method based on the generative model provided by the present invention, according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm for multiple iterations, we get multiple adjusted target transformation plans, including: According to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , calculate the dominance relationship between the objective functions, and obtain the non-dominated transformation scheme and the dominated transformation scheme; For the dominated renovation plans, according to the domination degree classification, the congestion degree is sorted according to the distances between the dominated renovation plans in each level; According to the degree of dominance and congestion of the dominated transformation plans, select The best transformation plan is used as the parent solution for the next generation; Calculate the crossover probability between parents Mutation probability , exchange or update the partial sequence of the transformation scheme, and also get The transformation plan of the next generation is used as the solution for the next generation; The parent solution of the next generation and the child solution of the next generation are used as the updated transformation plan, and processed by the first generative model and the second generative model to obtain the updated economic revenue and expenditure situation E after m years. m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , iterate continuously until the iteration round is met or the transformation solution set P is no longer updated; The result of the last iteration is output to obtain multiple adjusted non-dominated target transformation schemes.

[0013] According to the urban village reconstruction planning method based on the generative model provided by the present invention, the optimal target reconstruction plan is obtained by analyzing the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years, and the total urban internal commuting C within m years according to the multiple adjusted target reconstruction plans and their corresponding total economic revenue and expenditure E within m years, the total urban facility utilization I within m years, and the total urban internal commuting C within m years, specifically including: For each adjusted target transformation plan, its corresponding urban activity status g after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m Statistical processing is performed separately to obtain the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years, and the total internal urban commuting C within m years; The optimal target transformation plan is determined based on the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years, and the total internal urban commuting C within m years and their respective weights.

[0014] The present invention also provides a system for planning urban village reconstruction based on a generative model, comprising: The first parameter generation module is used to generate the urban activity status g m years later using the pre-trained first generative model based on the determined urban status g, urban village status u and transformation plan set P. m; Among them, the first generative model is trained based on the urban condition training samples, the urban village condition samples and the transformation plan sample set; The second parameter generation module is used to generate the urban activity status g m years later. m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; Iterative calculation module, used to obtain the economic revenue and expenditure E after m years based on the transformation plan set P m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations and obtain multiple adjusted target transformation plans; The target transformation plan generation module is used to analyze the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total internal urban commuting situation C within m years to obtain the optimal target transformation plan.

[0015] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the urban village renovation planning method based on the generative model as described in any one of the above-mentioned methods is implemented.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for planning the renovation of an urban village based on a generative model as described in any one of the above is implemented.

[0017] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for planning urban village reconstruction based on a generative model.

[0018] The method and system for planning urban village reconstruction based on a generative model provided by the present invention generate the urban activity status gm m years later using a pre-trained first generative model according to the determined urban status g, urban village status u and reconstruction plan set P, and then generate the urban activity status gm m years later according to the urban activity status g m years later.m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m , and then use the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations, obtain multiple adjusted target transformation plans, and then select the optimal target transformation plan. The method of this embodiment uses a generative model to quickly simulate and predict the future economic, social and environmental impacts of different transformation plans, so that multiple goals such as economic benefits, public service facility utilization and urban internal commuting can be considered at the same time, and the best balance point can be found through multi-objective optimization, providing decision makers with more comprehensive and accurate data support, so as to make more scientific and reasonable planning decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 It is one of the flow charts of the urban village reconstruction planning method based on the generative model provided by the present invention.

[0021] Figure 2 It is a schematic diagram of the framework of the urban village reconstruction planning method provided by the present invention.

[0022] Figure 3 It is a flowchart of the training method of the first network provided by the present invention.

[0023] Figure 4 It is a schematic diagram of the framework of the first generative model provided by the present invention.

[0024] Figure 5 It is a flow chart of the training method of the second network provided by the present invention.

[0025] Figure 6 This is the second flow chart of the urban village reconstruction planning method based on the generative model provided by the present invention.

[0026] Figure 7 It is a schematic diagram of the framework of the second generative model provided by the present invention.

[0027] Figure 8This is the third flow chart of the urban village renovation planning method based on the generative model provided by the present invention.

[0028] Fig. 9 It is a schematic diagram of the framework of the heuristic algorithm H provided by the present invention.

[0029] Fig.10 This is the fourth flow chart of the urban village reconstruction planning method based on the generative model provided by the present invention.

[0030] Fig.11 It is a schematic diagram of the structure of the urban village reconstruction planning system based on the generative model provided by the present invention.

[0031] Fig.12 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] First, the terms involved in this embodiment are schematically explained.

[0034] E (Economic income and expenditure): This indicator reflects the economic benefits of the renovation project. It may include profits from commercial housing development, rental income from rental housing, etc. Economic benefits are an important indicator for evaluating whether a renovation project is economically feasible. It involves the return on investment and economic benefits of the renovation project.

[0035] I (Urban Facility Utilization, i.e. Moran's Index): The Moran's Index is a spatial autocorrelation statistic that measures the degree of clustering of spatial data. Here, I refers to the utilization of public service facilities after transformation, which is evaluated by the Moran's Index. A higher Moran's Index indicates that the use of public service facilities is more spatially clustered, which may mean that the distribution and utilization of facilities are more efficient.

[0036] C (Internal commuting in the city): This indicator reflects the commuting conditions within the city after the transformation, which may include commuting time, commuting distance, traffic flow, etc. Internal commuting in the city is an important indicator for evaluating the efficiency of urban operations and the quality of life of residents, and it involves the planning and optimization of the transportation system.

[0037] Generative Model: A generative model is a type of machine learning model that can generate new data instances. They are usually based on probabilistic generation methods, and can learn the potential distribution of data from training data and generate new data samples accordingly. In the planning of urban village renovation, the application of generative models includes: Generative models can generate urban activity conditions at a certain point in the future based on the existing urban conditions, urban village conditions, and renovation plans. This includes economic activities, population distribution, traffic flow, etc.; then the model can predict the internal commuting conditions and access to urban facilities in the city after the implementation of the renovation plan, helping planners evaluate the potential impact of different renovation plans; by simulating different renovation plans, generative models can help planners understand the impact of these plans on the future development of the city, so as to optimize the plans.

[0038] Heuristic Algorithm: Heuristic algorithms are a class of algorithms used to solve optimization and search problems. They are usually based on experience and intuition to guide the search process to find approximate solutions to the problem. Heuristic algorithms are not necessarily guaranteed to find the global optimal solution, but in practical applications, they can often find a good enough solution in a reasonable time. Heuristic algorithms are used to process multiple adjusted transformation plans, optimize the plans through multiple iterations, and finally obtain the optimal target transformation plan. This involves calculating the dominance relationship between objective functions, crowding sorting, selecting the optimal solution as the parent, and generating new child solutions through crossover and mutation operations.

[0039] Non-dominated solution set: The non-dominated solution set contains solutions that are optimal in all objectives. These solutions represent trade-offs between different objectives. In the non-dominated solution set, no solution is strictly better than other solutions in all objectives. For example, consider an optimization problem with two objectives: minimize cost and minimize time. Suppose we have three solutions A, B, and C, and their objective function values ​​are as follows: A: cost = 10, time = 5; B: cost = 8, time = 6; C: cost = 9, time = 4. In this example: B is dominated by A and C because A has better cost and time than B, while C has better time than B. Neither A nor C dominates each other because they have a trade-off between objectives: A has better cost than C, but C has better time than A. Therefore, the non-dominated solution set is {A,C}.

[0040] Combine the following Figure 1-Figure 10 The invention describes a method and system for planning urban village reconstruction based on a generative model according to an embodiment of the invention.

[0041] In this embodiment, for a given urban area, it is divided into A fixed range spatial grid, , where each grid has a corresponding set of attributes , which respectively correspond to the size of the residential population, second-hand housing transaction prices, rental housing prices, and the number of facilities organized by type within the city grid.

[0042] On this basis, we should identify the areas to be transformed in the city. The spatial range of urban villages is expressed as For each urban village, it has a set of core attributes: , including corresponding grid coordinates, land area, building volume ratio, and number of facilities organized by type.

[0043] Planning tasks are to be planned within the year range Inside, for Each urban village to be planned has a corresponding development plan to form a complete set of planning plans Each planning scheme includes the following information , which correspond to the main indicators of urban village renovation, such as the proportion of urban village renovation area, the proportion of commercial housing development, the proportion of rental housing development, the proportion of public facilities increase, and the volume ratio. Since there is no reverse process in urban village renovation, there is In addition, external parameters include the expected growth rate of housing prices. The average annual inflation rate .

[0044] The goal of the plan is to achieve the goal of "maintaining growth, making up for shortcomings, and adjusting the structure" within the city as a whole. Among them, "maintaining growth" refers to the economic revenue and expenditure before and after the transformation, which is composed of the profits brought by commercial housing development and the rent provided by leasing housing, of which the accumulated The annual economic benefits can be calculated by the following formula (1): (1)

[0045] Where n: represents the index of each plot or area in the urban village.

[0046] ∑n: means to sum all plots or areas.

[0047] Area n : Represents the area of ​​the nth plot.

[0048] : Represents the transformation completion rate of the nth plot after m years.

[0049] : Represents the economic development rate after m years.

[0050] : Represents the floor area ratio of the nth plot m years later.

[0051] : Represents the unit price in economic activities, such as real estate prices.

[0052] η1 and η2: represent parameters related to economic development and inflation rate.

[0053] : Represents the rental rate after m years.

[0054] : represents the continuous product from x=0 to x=30.

[0055] “Filling the gaps” refers to the utilization of public service facilities after transformation. The Moran index of the number of visits to public service facilities between each grid can be calculated by the following formula (2): (2)

[0056] Among them, i and j: These are usually the dimensions of the spatial grid, representing the number of grids the city is divided into. For example, if the city is divided into a grid with i rows and j columns, then i×j is the total number of grids.

[0057] : These are summation symbols, meaning the sum is taken over all grids.

[0058] : This is a weight matrix, where Represents the weight between mesh a and mesh b. This weight may be based on the distance, connectivity, or other relevant factors between the meshes.

[0059] and : These represent the Point of Interest vectors in grid a and grid b. These vectors may contain information about the facilities in that grid, such as the type of facility, quantity, or frequency of use.

[0060] : This represents the utilization of urban facilities after m years. It may be a vector containing the utilization of facilities in each grid.

[0061] : This represents the average vector of the points of interest across all grids.

[0062] “Structural adjustment” refers to the commuting situation within the city after the transformation, which is calculated by the total commuting volume between grids, see the following formula (3): (3)

[0063] Note that after the transformation, the population distribution within the city 、Facility layout The factors involved in such situations are too complicated. It is not possible to export directly from the transformation plan.

[0064] Therefore, the core goal of the present invention is to train the first generative model , can be based on the existing urban conditions to come up with specific transformation plans The city’s condition after 2017 is analyzed, and the above formula is used to further calculate the three main indicators under this scenario.

[0065]

[0066] Then, by training the second generative model , can be based on the transformed urban conditions ,predict the urban facility access and internal commuting in this scenario.

[0067]

[0068] Finally, through the heuristic algorithm model , adjust the transformation plan In order to maximize the overall benefits of the plan, as a plan for urban village reconstruction:

[0069] Specifically, see Figure 1 and Figure 2 . Figure 1 This is one of the flow charts of the urban village reconstruction planning method based on the generative model provided by the present invention. Figure 2 It is a schematic diagram of the framework of the urban village reconstruction planning method provided by an embodiment of the present invention.

[0070] The method includes the following: Step 101: Generate the urban activity status g m years later using the pre-trained first generative model based on the determined urban status g, urban village status u and transformation plan set P m .

[0071] Among them, the first generative model is trained based on urban condition training samples, urban village condition samples and transformation plan sample sets.

[0072] Specifically, data collection and preprocessing include: City status data: Collect detailed data on the current status of the city, including demographic data, land use types, traffic flow, economic activity data, etc. This data can be obtained from sources such as municipal departments, transportation management bureaus, and statistical bureaus.

[0073] Urban Village Condition Data: Collect detailed condition data of urban villages, including building density, infrastructure conditions, living conditions, environmental sanitation, etc. These data can be obtained through on-site surveys, satellite remote sensing, drone photography, etc.

[0074] Transformation plan set P: Collect or formulate a transformation plan set, including planned construction projects, infrastructure improvement plans, estimated investment amounts, etc. These plans can be provided by urban planning departments or formulated based on expert consultation and public participation.

[0075] Integrate the collected data into a unified data format for input into the generative model. Preprocess the data, including normalization, removal of outliers, and filling in missing data, to improve the prediction accuracy of the model.

[0076] Load the pre-trained first generative model GenAI α (·) and set the model parameters, including the predicted time range m years. The preprocessed urban conditions, urban village conditions and transformation plan set are input into the model. The model generates the urban activity status g m years later based on the input data and the learned patterns. m . The predicted urban activity status g m It is applied to urban planning, transportation planning, infrastructure layout and other fields to provide scientific basis for decision makers.

[0077] Step 101 can provide data-driven future forecasts for urban village renovation planning, help planners evaluate the potential impacts of different renovation plans, and provide a scientific basis for subsequent planning decisions.

[0078] Step 102: Based on the city activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m .

[0079] It should be noted that generative models can take many forms, for example, neural network models, such as generative adversarial networks (GANs), variational autoencoders (VAEs), autoregressive models, etc. These models are based on neural network architecture and can learn complex distributions of data and generate new data instances; probability-based models: such as hidden Markov models (HMMs), Bayesian networks, etc. These models simulate the distribution of data by defining the probabilistic process of data generation; optimization-based models: such as some energy-based models (EBMs), which learn the distribution of data by optimizing the energy function.

[0080] In this embodiment, the first generative model is a neural network model, and the second generative model is a Markov generative model.

[0081] In this embodiment, based on the urban activity status g after m years m , extract relevant urban facility data, including facility type, location, capacity, etc. Collect information such as demographic data, employment status, and living habits of urban residents, which will serve as the attributes of the agent. Define the agent's behavioral parameters for the second generative model, including exploration probability, return probability, and movement cost. These parameters can be determined through statistical analysis of historical data, expert consultation, or pre-experiments.

[0082] Initialize the position of the agent randomly or according to specific rules in the spatial grid of the city. Assign an initial state to each agent, including the grid it is in, the city facilities it has visited, etc. Then use the second generative model to simulate the agent's movement behavior in the city, and each agent moves between the city grids according to the defined parameters. The movement of the agent may be affected by factors such as the distribution of urban facilities, transportation networks, and personal preferences.

[0083] According to the agent's movement trajectory and access records, the number of visits to each urban facility is counted. The calculation method built into the second generative model is used to estimate the utilization of urban facilities m years later. m .

[0084] For the generated I m The results are evaluated to check whether they meet the expected urban development goals and planning requirements. According to the evaluation results, the model parameters are adjusted or the model structure is optimized to improve the accuracy and reliability of the prediction.

[0085] Through step 102, the second generative model It can provide urban planners with in-depth insights into the future utilization of urban facilities and help them formulate more scientific and reasonable urban village renovation plans.

[0086] Step 103: Obtain the economic revenue and expenditure situation E after m years based on the transformation plan set P. m, according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m ,The heuristic algorithm is used for multiple iterations to obtain multiple adjusted target transformation plans.

[0087] Initialize the transformation plan set: Starting from the data obtained in step 101 and step 102, initialize the transformation plan set P, including all possible transformation plans and their corresponding economic, commuting and facility utilization prediction data.

[0088] Define evaluation indicators: Determine the indicators for evaluating the transformation plan, including but not limited to economic benefits, commuting efficiency, facility utilization, etc. Assign weights to each indicator to reflect its importance in the overall plan.

[0089] In this embodiment, a suitable heuristic algorithm, such as genetic algorithm, simulated annealing, particle swarm optimization, etc., can be selected to search for the optimal transformation solution. The parameters of the algorithm are set, such as population size, number of iterations, crossover rate, and mutation rate.

[0090] A heuristic algorithm is used to iteratively optimize the transformation solution set P.

[0091] 1) In each generation, evaluate the fitness of each solution, that is, its performance on the evaluation index. Select the best solution as the parent based on the fitness, and generate new child solutions through crossover and mutation operations.

[0092] 2) Perform non-dominated sorting on the population of each generation and identify non-dominated transformation schemes. For the dominated schemes, calculate their crowding degree to maintain the diversity of the population.

[0093] 3) According to the non-dominated sorting and crowding selection, the parent solution of the next generation is selected. The selected parent solution and the generated child solution form a new transformation solution set for the next round of iteration.

[0094] Repeat steps 1) to 3) until the preset number of iterations is met or the transformation solution set is no longer significantly updated.

[0095] Evaluate the optimal solution set of each generation and monitor the convergence performance of the algorithm.

[0096] After the final iteration, the non-dominated solution sets within the population are output, which represent the transformation planning schemes that achieve the best balance between multiple objectives. The optimal transformation scheme is provided to the decision maker to support him in making the final planning decision.

[0097] Step 103 provides a systematic framework for optimizing the urban village renovation planning scheme through a heuristic algorithm to ensure that the scheme achieves the best balance among multiple goals such as economy, society and environment.

[0098] Step 104: Analyze the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure E within m years, total urban facility utilization I within m years, and total internal urban commuting C within m years to obtain the optimal target transformation plan.

[0099] In this embodiment, for each adjusted target transformation plan, the corresponding urban activity status g after m years is m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m Statistical processing is performed separately to obtain the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years and the total urban internal commuting C within m years. Then, the optimal target transformation plan is determined based on the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years and the total urban internal commuting C within m years and their corresponding weights.

[0100] The weighted sum method is used to calculate the comprehensive score of each renovation plan. For example, for each plan P i , its comprehensive score S i It can be expressed as formula (4): Si=w E ·E i +w I I i +w C ·C i (4)

[0101] Among them, w E 、w I 、w C are the weights of economic, facility utilization and commuting indicators, E i ,I i , C i It is Plan P i scores on these indicators.

[0102] Step 104 provides a comprehensive analysis and decision-making framework to help decision makers select the best target transformation plan from multiple adjusted transformation plans and provide support for the implementation and subsequent evaluation of the plan.

[0103] The urban village reconstruction planning method based on the generative model provided in the embodiment of the present invention generates the urban activity status g m years later by using the pre-trained first generative model according to the determined urban status g, urban village status u and reconstruction plan set P m , and then according to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m , and then use the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations, obtain multiple adjusted target transformation plans, and then select the optimal target transformation plan. The method of this embodiment uses a generative model to quickly simulate and predict the future economic, social and environmental impacts of different transformation plans, so that multiple goals such as economic benefits, public service facility utilization and urban internal commuting can be considered at the same time, and the best balance point can be found through multi-objective optimization, providing decision makers with more comprehensive and accurate data support, so as to make more scientific and reasonable planning decisions.

[0104] Furthermore, the first generative model and the second generative model involved in this embodiment are schematically illustrated.

[0105] The first generative model includes a first network and a second network, wherein the first network is U-Net and the second network is ResNet.

[0106] The essence of the generative model is to fit the given data sample through a known probability model, so as to extract samples from the known distribution and obtain a new instance consistent with the given data sample. The diffusion generative model is a branch of the generative model. Its basic idea is to gradually add noise to the data through the diffusion process. When the noise is large enough, the sample is destroyed into noise (the distribution is known). By fitting this diffusion process through the first network, the noise added at each step can be predicted.

[0107] In the denoising process, we start with random noise and use the trained first network to predict the noise added to the image. By removing the noise, we gradually restore the noise-free image. At the same time, in order to guide the denoising process so that it can correctly generate the urban village situation U and the urban form under the given transformation plan P, we also need to train a corresponding second network to select the correct direction of noise removal.

[0108] See also Figure 3 and Figure 4, the training method of the first network includes: 301. Randomly select a city condition training sample from the training sample data set .

[0109] Randomly select a sample from the pre-collected urban condition training sample dataset . Ensure that the sample g0 contains sufficient historical information to represent the state of the city at a specific point in time.

[0110] 302. Determine the random extraction time t and select from a constant sequence Extract the corresponding constant .

[0111] Determine a random extraction time point, which can be a time point selected based on a certain strategy (such as uniform distribution, specific event trigger, etc.). Take a constant corresponding to the extraction time from a predefined constant sequence, which may represent the influence weight of time or other time-related coefficients.

[0112] Randomly select a time interval m, which defines the time range from the current time point forward to the beginning of the sample data set.

[0113] 303. From a training sample with urban conditions A random sample noise is randomly drawn from a standard Gaussian distribution of the same size , and the random sample noise And constant Add to urban condition training samples In the above example, we get the urban condition training samples after adding noise. .

[0114] Among them, the urban condition training samples after adding noise Calculated by the following formula (5): (5)

[0115] 304. The noise-added urban condition training sample The time of random selection Input into the first network to get the predicted noise .

[0116] 305. Calculate a loss value according to the predicted noise and the random sample noise, and update parameters in the first network according to the predicted loss value by using a stochastic gradient descent method until the parameters in the first network converge.

[0117] The loss function is

[0118] Through steps 301 to 305, the first network of the first generative model can learn to predict noise from the noisy urban conditions and gradually restore the original urban activity conditions. This training method helps to improve the performance of the model in the denoising process, thereby generating more accurate future urban condition predictions and providing more reliable data support for urban village reconstruction planning.

[0119] See also Figure 4 and Figure 5 , the training method of the second network includes: 501. Randomly select a city condition training sample from the training sample data set .

[0120] 502. From randomly selected urban condition training samples The corresponding time point is taken as the starting point, and the time point forward to the beginning of the sample data set is taken as the ending point, and a time interval m is randomly selected.

[0121] 503. According to the extracted time interval m, obtain the known urban village condition sample U0 and the corresponding transformation plan sample set P θ .

[0122] 504. The randomly selected urban condition training samples Input into the second network to get the predicted transformation time interval 、The situation of urban villages And transformation plan P θ .

[0123] 505. Calculate the predicted renovation time interval The time interval between the extraction and the predicted urban village status Compared with the known urban village condition samples and the predicted transformation plan P θ The difference between the known transformation solution sample and the second loss value is obtained.

[0124] Among them, the loss function is: .

[0125] 506. Update the parameters of the second network using the second loss value through a stochastic gradient descent method until the parameters of the second network converge.

[0126] The second network of the first generative model can learn to predict the future transformation effect from the current state of the city and optimize the model parameters based on historical data, thereby improving the accuracy of the model in predicting future urban activities. This training method helps to generate more accurate predictions of future urban conditions and provide more reliable data support for urban village transformation planning.

[0127] See also Figure 6, step 101 specifically includes: 601. According to the determined urban condition g, extract the noise amount from the standard Gaussian distribution, and then add it to the determined urban condition g to obtain the current random urban condition .

[0128] 602. Random city status with a predefined maximum number of denoising steps Input into the pre-trained first network for processing to obtain the predicted noise .

[0129] 603. Random city status , the urban village status u and the transformation plan set P are input into the pre-trained second network to calculate the gradient value of the loss function .

[0130] 604. Update the noise according to the gradient value , and update the current random city state according to the updated noise to obtain the updated random city state.

[0131] The updated noise is calculated by the following formula (6): (6)

[0132] The updated random city status is calculated by the following formula (7): (7)

[0133] 605. Iteratively execute the step of inputting the updated random city status into the first network and the second network respectively to obtain the next updated random city status until the maximum number of denoising steps T is 0, and obtain the city activity status m years later.

[0134] Through the above steps, we only need to specify the maximum number of denoising steps and constant sequence , we can get the first generative model , there are two related first networks and the second network Together, they can be combined to form a 、The situation of urban villages And the given transformation plan ,generate The city's situation after .

[0135] In addition, for the second generative model , can be based on the transformed urban conditions m, predicting the city facility visits in this scenario and internal commuting The general framework can be found in Figure 7 .

[0136]

[0137] The core of the problem is that the factors that affect commuting in the city are very complex, and the commuting conditions in the city are difficult to predict as a whole. In this patent, we use a method based on exploration-memory-mobility behavior to generate the behavior of each micro-individual in the city and sum them up to obtain the facility access and commuting conditions of the entire city.

[0138] According to existing research, human movement in cities is controlled by a power law distribution, that is, there are fewer long-distance movements and more short-distance movements, a small number of destinations occupy most of the trips, and most destinations have only a small number of trips. The reason for this phenomenon is that human activities in cities are memorized (affected by previous activities) and social (affected by other human activities). Therefore, we adopt a two-stage approach, taking into account both the situation of humans returning to previously visited places and other humans. The basic process of generation is as follows Figure 8 As shown: 801. Define multiple agents in a city. For any agent in the city , recording the set of all places it has visited.

[0139] 802. For any time , pick a random number from a random distribution .

[0140] 803. Calculate the exploration probability of agent i .

[0141] in, and are two predetermined hyperparameters, is the number of grid points that agent i has visited.

[0142] 804. If p ≥ The control agent performs a return operation, randomly selects a location from the set of locations it has visited, and updates its location.

[0143] The probability of visiting each location is proportional to its previous visit frequency: .

[0144] 805 If , control agent i to perform exploration operations, randomly select a location from the set of other locations in the city, and update its location; among them, the visit probability of each location is proportional to the number of people at the location, and the visit probability of each location is inversely proportional to the distance between two locations.

[0145] Specifically, see the following formula (8): (8)

[0146] in, They are the starting location and random location respectively; is the distance between two locations; is the core parameter.

[0147] 806. Perform the above steps for all agents in the city until the time Reach the specified time step .

[0148] 807. Count the number of visits to the city facilities of interest and calculate the utilization of the city facilities m years later

[0149] 808. Count the number of passes between grids within the city and calculate the commuting situation within the city m years later .

[0150] Note that in this process, the core parameters The parameters of the observed samples are unknown and may vary between individuals. This patent constructs a Markov generative model to generate possible samples by solving the parameters that make the observed samples most likely to appear from a part of the observed human movement process through the maximum likelihood method.

[0151] Assume the core parameters of the model All come from standard Gaussian distribution .

[0152] According to the existing human behavior trajectory sequence , count the exploration behavior frequency, return behavior frequency and visited location data of all samples, and calculate the core parameters of each sample ; Based on the statistically obtained parameters, calculate the log-likelihood function that makes this series of samples appear

[0153] Using the obtained probability distribution , according to the given urban conditions , extract and generate a certain proportion of agents for each grid, and execute the above steps 801-808.

[0154] Alternatively, see Fig. 9and Fig.10 . Fig. 9 It is a schematic diagram of the framework of the heuristic algorithm H. Fig.10 1 is a flow chart of step 103. Specifically, step 103 includes: 1001. Based on the economic revenue and expenditure situation m years later, E m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , calculate the dominance relationship between the objective functions, and obtain the non-dominated transformation scheme and the dominated transformation scheme.

[0155] Before executing step 1001, the relevant parameters of the transformation plan , both upper and lower bounds are set by experts and . Set the population size , iteration round , crossover probability , mutation probability ; Random initialization Different transformation planning schemes are obtained, and the solution set .

[0156] 1002. For the dominated transformation plans, according to the grading of the domination degree, the congestion degree is sorted according to the distance between the dominated transformation plans in each level.

[0157] 1003. According to the degree of control and congestion of the controlled transformation plans, select The optimal transformation plan will be used as the paternal solution for the next generation.

[0158] 1004. Calculate the crossover probability between fathers Mutation probability , exchange or update the partial sequence of the transformation scheme, and also get The transformation plan for the next generation will be used as the solution for the next generation.

[0159] The offspring solutions generated by crossover and mutation operations are not guaranteed to be "optimal". They are new candidate solutions that need to be evaluated together with other individuals in the selection operation of the next generation. Only those offspring solutions that perform well in the evaluation will be selected as the parents of the next generation. This process helps the algorithm explore in the search space and gradually approach the true Pareto optimal solution set.

[0160] 1005. The next generation's parent solution and the next generation's child solution are used as the updated transformation plan, and are processed by the first generative model and the second generative model to obtain the updated economic revenue and expenditure situation E after m years. m, the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , and iterates continuously until the iteration round is met or the transformation solution set P is no longer updated.

[0161] 1006. Output the result of the last iteration to obtain multiple adjusted non-dominated target transformation plans.

[0162] In each generation of the genetic algorithm, a new generation of population is generated through selection, crossover and mutation operations, including k parent solutions and k child solutions. These solutions represent the best transformation planning solutions in the current population, and they are solutions that balance multiple objectives.

[0163] The above heuristic algorithm model , search for the optimal set of transformation solutions for that year in the given solution space. Usually, we will not get a single solution, but a set of solutions with advantages and disadvantages in different directions. These solutions are all feasible and excellent transformation plans, with different preferences for different transformation goals, which can be compared and selected by decision makers in the actual process.

[0164] Taking City A as an example, if we want to generate an adaptive planning scheme for the urban village reconstruction in City A, we can achieve this through the following methods: (1) City A was divided into spatial grids with a grid size of 1,200 meters and a distance of 99 and 125 in the longitude and latitude directions, respectively. Non-urban areas were removed, resulting in a total of 2,267 grids.

[0165] (2) The urban village plots in City A were extracted, and a total of 1,481 urban villages to be renovated and related information were obtained.

[0166] (3) Use historical data on urban village renovation in places A, B, and C to train a generative model , where a sufficiently long time is set , ,in .in, is a time-dependent parameter that controls the noise level at each step in the denoising process; is another time-dependent parameter, which determines the proportion of noise removed in each step; T represents the maximum number of steps in the denoising process, which is a hyperparameter that can be adjusted according to the complexity of the model and the characteristics of the training data.

[0167] (4) Using the crowd behavior activity sequence of City A to train the generative model ,In order to speed up the calculation, the ratio of the model-generated agents to the real population is 1:1000.

[0168] (5) Set the parameters required for operation, where the solution space is set as follows: 0.6 < r a < 1, 0.5 < r e <0.8, 0.15 < r r <0.25, 1.3 < r poi <1.5, 1.8 < FAR < 2.6; the total renovation years m = 10, η1 = 1.05, η2 = 1.02 are set for the external parameters; the internal parameters of the heuristic model are k = 50, g = 10, p c = 0.8, p m = 0.5.

[0169] 0.6 < r a <1: represents the range of the renovation completion rate r a This ratio may reflect the proportion of urban villages that have completed renovation within a given time period.

[0170] 0.5 < r e <0.8: represents the range of the economic development rate r e This ratio may be related to the urban economic growth or the development speed of a specific area.

[0171] 0.15 < r r <0.25: represents the range of the rent rate r r This ratio may reflect the change in the rent level of the renovation area.

[0172] 1.3 < r poi <1.5: represents the range of the urban facility utilization r poi This ratio may be related to the usage frequency or efficiency of urban facilities.

[0173] 1.8 < FAR < 2.6: represents the range of the floor area ratio FAR. The floor area ratio is a measure of building density and reflects the total amount of buildings allowed on a given land area.

[0174] (6) According to the method described above, use the heuristic model , call the trained first generative model and the second generative model , and obtain 4 non-dominated renovation sets

[0175] In this embodiment, there are 3 evaluation dimensions (development benefit, facility configuration, traffic evaluation). Therefore, select the best renovation plan set in each dimension, and add a renovation plan set that balances between various goals. So there are 4 non-dominated renovation plan sets.

[0176] (7) Analysis of four non-dominated transformation scheme sets respective According to the comparison between their scores, we give them corresponding names and conduct analysis.

[0177] By selecting the optimal transformation plan, limited financial and construction resources can be allocated more effectively and resource utilization efficiency can be improved. Choosing a plan that takes facility configuration and commuting conditions into consideration can enhance the overall service function of the city and improve the quality of life of residents.

[0178] The following is a description of the urban village renovation planning system based on a generative model provided in an embodiment of the present invention. The urban village renovation planning system based on a generative model described below and the urban village renovation planning method based on a generative model described above can be referenced to each other.

[0179] The embodiment of the present invention provides a system for planning urban village reconstruction based on a generative model. Fig.11 ,include: The first parameter generation module 111 is used to generate the urban activity status g m years later using the pre-trained first generative model according to the determined urban status g, urban village status u and transformation plan set P. m ; Among them, the first generative model is trained based on the urban condition training samples, the urban village condition samples and the transformation plan sample set; The second parameter generation module 112 is used to generate a parameter based on the urban activity status g after m years. m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; Iterative calculation module 113, used to obtain the economic income and expenditure situation E after m years according to the transformation plan set P m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations and obtain multiple adjusted target transformation plans; The target transformation plan generation module 114 is used to analyze the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure E within m years, total urban facility utilization I within m years, and total internal urban commuting C within m years to obtain the optimal target transformation plan.

[0180] The system of this embodiment uses a generative model to quickly simulate and predict the future economic, social and environmental impacts of different transformation plans, so that it can simultaneously consider multiple goals such as economic benefits, public service facility utilization and inner-city commuting, and find the best balance point through multi-objective optimization, providing decision makers with more comprehensive and accurate data support, so as to make more scientific and reasonable planning decisions.

[0181] Fig.12 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.12 As shown, the electronic device may include: a processor 1210, a communication interface 1220, a memory 1230 and a communication bus 1240, wherein the processor 1210, the communication interface 1220 and the memory 1230 communicate with each other through the communication bus 1240. The processor 1210 may call the logic instructions in the memory 1230 to execute the urban village reconstruction planning method based on the generative model, the method comprising: according to the determined urban condition g, the urban village condition u and the reconstruction plan set P, using the pre-trained first generative model, generate the urban activity condition g after m years m ; Among them, the first generative model is trained based on the urban condition training samples, urban village condition samples and transformation plan sample sets; according to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; According to the transformation plan set P, the economic revenue and expenditure situation m years later is obtained E m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m, using the heuristic algorithm to perform multiple iterations to obtain multiple adjusted target transformation plans; analyzing the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total urban internal commuting situation C within m years to obtain the optimal target transformation plan.

[0182] In addition, the logic instructions in the above-mentioned memory 1230 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0183] On the other hand, the present invention also provides a computer program product, the computer program product includes a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the urban village reconstruction planning method based on the generative model provided by the above methods, the method comprising: according to the determined urban condition g, the urban village condition u and the reconstruction plan set P, using the pre-trained first generative model to generate the urban activity condition g after m years m ; Among them, the first generative model is trained based on the urban condition training samples, urban village condition samples and transformation plan sample sets; according to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; According to the transformation plan set P, the economic revenue and expenditure situation m years later is obtained E m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later Cm And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations to obtain multiple adjusted target transformation plans; analyzing the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total urban internal commuting situation C within m years to obtain the optimal target transformation plan.

[0184] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the urban village reconstruction planning method based on the generative model provided by the above methods, the method comprising: according to the determined urban condition g, the urban village condition u and the reconstruction plan set P, using the pre-trained first generative model to generate the urban activity condition g m years later m ; Among them, the first generative model is trained based on the urban condition training samples, urban village condition samples and transformation plan sample sets; according to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; According to the transformation plan set P, the economic revenue and expenditure situation m years later is obtained E m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations to obtain multiple adjusted target transformation plans; analyzing the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total urban internal commuting situation C within m years to obtain the optimal target transformation plan.

[0185] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0186] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for planning urban village reconstruction based on a generative model, characterized in that: include: Based on the determined urban conditions g, urban village conditions u, and transformation plan set P, the pre-trained first generative model is used to generate the urban activity conditions g m years later. m ; Among them, the first generative model is trained based on the urban condition training samples, the urban village condition samples and the transformation plan sample set; According to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; According to the transformation plan set P, the economic revenue and expenditure situation E after m years is obtained m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations and obtain multiple adjusted target transformation plans; The optimal target transformation plan is obtained by analyzing the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total internal urban commuting situation C within m years.

2. The urban village reconstruction planning method based on the generative model according to claim 1 is characterized in that: The first generative model includes a first network and a second network; Based on the determined urban conditions g, urban village conditions u, and transformation plan set P, the pre-trained first generative model is used to generate the urban activity conditions g m years later. m , specifically including: According to the determined urban condition g, the noise amount is extracted from the standard Gaussian distribution, and then added to the determined urban condition g to obtain the current random urban condition ; Randomize city status with a predefined maximum number of denoising steps Input into the pre-trained first network for processing to obtain the predicted noise ; Randomize city status , the urban village status u and the transformation plan set P are input into the pre-trained second network to calculate the gradient value of the loss function ; Update the noise according to the gradient value , and update the current random city status according to the updated noise to obtain the updated random city status; Iterative execution inputs the updated random city status into the first network and the second network respectively to obtain the next updated random city status step until the maximum number of denoising steps T is 0, and the city activity status m years later is obtained.

3. The urban village reconstruction planning method based on the generative model according to claim 2 is characterized in that: The training method of the first network includes: Randomly select a city condition training sample from the training sample data set; Determine the random draw time , and from a constant sequence Extract the corresponding constant ; Randomly extracting a random sample noise from a standard Gaussian distribution of the same size as the urban condition training sample, and adding the random sample noise and the constant to the urban condition training sample to obtain a noisy urban condition training sample; The noised urban condition training samples are compared with the randomly selected time Input into the first network to obtain the predicted noise; A loss value is calculated according to the predicted noise and the random sample noise, and the parameters in the first network are updated according to the predicted loss value by using the stochastic gradient descent method until the parameters in the first network converge.

4. The urban village reconstruction planning method based on the generative model according to claim 2 is characterized in that: The training method of the second network includes: Randomly select a city condition training sample from the training sample data set ; From randomly selected urban condition training samples The corresponding time point is taken as the starting point, and the time point forward to the beginning of the sample data set is taken as the end point, and a time interval is randomly selected ; According to the extraction time interval , obtain known samples of urban village conditions And the corresponding transformation plan sample set ; The randomly selected urban condition training samples Input into the second network to get the predicted transformation time interval 、The situation of urban villages and transformation plans ; Calculate the predicted renovation time interval The time interval between the extraction and the predicted urban village status Compared with known samples of urban villages and predicted transformation plans The difference between the sample and the known transformation plan is used to obtain the second loss value; By using the stochastic gradient descent method, the parameters of the second network are updated according to the second loss value until the parameters of the second network converge.

5. The urban village reconstruction planning method based on a generative model according to claim 1 is characterized in that: According to the urban activity status g after m years m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m , specifically including: Define multiple agents in a city. For any agent in the city , records the set of all places it has visited; For any time , pick a random number from a random distribution ; Calculate the exploration probability of agent i ,in, and are two predetermined hyperparameters, is the number of grid points that agent i has visited; If p≥ The control agent performs a return operation, randomly selects a location from the set of locations it has visited, and updates its position; the probability of visiting each location is proportional to the frequency of its previous visits; like , control agent i to perform exploration operations, randomly select a location from the set of other locations in the city, and update its location; the visit probability of each location is proportional to the number of people at the location, and the visit probability of each location is inversely proportional to the distance between two locations; Repeat the above steps for all agents in the city until time Reach the specified time step ; Count the number of visits to the city facilities of interest and calculate the utilization of the city facilities m years later Count the number of passes between grids within the city and calculate the commuting situation within the city m years later .

6. The urban village reconstruction planning method based on a generative model according to claim 1 is characterized in that: According to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm for multiple iterations, we get multiple adjusted target transformation plans, including: According to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , calculate the dominance relationship between the objective functions, and obtain the non-dominated transformation scheme and the dominated transformation scheme; For the dominated renovation plans, according to the domination degree classification, the congestion degree is sorted according to the distances between the dominated renovation plans in each level; According to the degree of dominance and congestion of the dominated transformation plans, select The best transformation plan is used as the parent solution for the next generation; Calculate the crossover probability between parents Mutation probability , exchange or update the partial sequence of the transformation scheme, and also get The transformation plan of each offspring is used as the offspring solution of the next generation; The parent solution of the next generation and the child solution of the next generation are used as the updated transformation plan, and processed by the first generative model and the second generative model to obtain the updated economic revenue and expenditure situation E after m years. m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , iterate continuously until the iteration round is met or the transformation solution set P is no longer updated; The result of the last iteration is output to obtain multiple adjusted non-dominated target transformation schemes.

7. The urban village reconstruction planning method based on a generative model according to claim 1 is characterized in that: According to the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years and total urban internal commuting situation C within m years, an optimal target transformation plan is obtained, which specifically includes: For each adjusted target transformation plan, its corresponding urban activity status g after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m Statistical processing is performed separately to obtain the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years, and the total internal urban commuting C within m years; The optimal target transformation plan is determined based on the total economic revenue and expenditure E within m years, the total urban facility utilization I within m years, and the total internal urban commuting C within m years and their respective weights.

8. A planning system for urban village reconstruction based on a generative model, characterized in that: include: The first parameter generation module is used to generate the urban activity status g m years later using the pre-trained first generative model based on the determined urban status g, urban village status u and transformation plan set P. m ; Among them, the first generative model is trained based on the urban condition training samples, the urban village condition samples and the transformation plan sample set; The second parameter generation module is used to generate the urban activity status g m years later. m , using the preset second generative model, calculate the inner city commuting situation C m years later m And the utilization of urban facilities after m years m Among them, the second generative model assumes that multiple agents move in the city and solves the movement parameters of multiple agents in the city through the maximum likelihood method, so as to predict the internal commuting situation C in the city m years after the transformation according to the transformed city conditions. m And the utilization of urban facilities after m years m ; Iterative calculation module, used to obtain the economic revenue and expenditure E after m years based on the transformation plan set P m , according to the economic revenue and expenditure situation E after m years m , the inner city commuting situation m years later C m And the utilization of urban facilities after m years m , using the heuristic algorithm to perform multiple iterations and obtain multiple adjusted target transformation plans; The target transformation plan generation module is used to analyze the multiple adjusted target transformation plans and their corresponding total economic revenue and expenditure situation E within m years, total urban facility utilization situation I within m years, and total internal urban commuting situation C within m years to obtain the optimal target transformation plan.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the urban village renovation planning method based on the generative model as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the urban village renovation planning method based on a generative model as described in any one of claims 1 to 7.

Images