Artificial intelligence driven urban system computational simulation platform method and related devices

Abstract

The present application discloses an artificial intelligence-driven urban system computational simulation platform method and related devices, relating to the fields of artificial intelligence and digital simulation technology. The method comprises: obtaining basic urban data to generate multiple initial strategies, then sampling to obtain a first reference strategy; deducing the first reference strategy through a strategy calculation model to obtain a first benchmark deduction result, and adding the result to a strategy recommendation set; screening at least two initial strategies through a target agent model to obtain a second reference strategy; deducing each second reference strategy through a strategy calculation model to obtain a second benchmark deduction result, and updating the strategy recommendation set based on the second reference strategy and the second benchmark deduction result to recommend strategies to users. The present application can implement urban system strategy recommendations and meet user control needs.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence and digital simulation technology, and in particular to an artificial intelligence-driven urban system computing simulation platform method and related devices. Background Art

[0002] The operation and development status of the urban system is mainly composed of the dynamics of multiple subsystems such as land use evolution and transportation. The various elements, flows, functions, facilities, and environments are interconnected and interact with each other, supporting the life processes of the urban system such as growth, decline (contraction), illness, renewal, and adaptation.

[0003] Traditional urban system simulation technologies (such as the UrbanSim model) focus on key sectors in urban development (households, businesses, developers, etc.), conducting dynamic simulations on an annual basis, from short to long-term perspectives. These simulations are primarily used to predict long-term patterns of land use and transportation under various regulatory scenarios. However, the core functional modules of current urban system simulation technologies focus primarily on simulation analysis rather than prediction. These simulations mostly use the BAU (Business-as-Usual) scenario, or baseline scenario, which depicts the natural state of urban elements developing according to existing trends without additional policy intervention. This makes the predictions difficult to effectively assess and predict the specific effects of policy implementation and the dynamic changes in urban systems in terms of policy responsiveness, thus limiting their application value in policy formulation and evaluation.

[0004] Therefore, how to recommend regulatory strategies for urban systems has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an artificial intelligence-driven urban system computational simulation platform method, apparatus, equipment, and medium that can recommend strategies for urban systems and meet user control needs. Furthermore, it can effectively evaluate and predict the specific effects of implemented strategies and the dynamic changes in urban systems.

[0006] To achieve the above objectives, a first aspect of an embodiment of the present application proposes an artificial intelligence-driven urban system computing simulation platform method, the method comprising:

[0007] Acquiring city basic data for city system calculation, performing simulation calculation on the city basic data, and obtaining city calculation element data;

[0008] generating at least two initial strategies based on the city calculation element data;

[0009] Sampling the at least two initial strategies to obtain at least two first reference strategies;

[0010] Deducing each of the first reference strategies using a pre-trained strategy calculation model to obtain a first benchmark deduction result, and adding the first reference strategy and the first benchmark deduction result to a strategy recommendation set;

[0011] Training the initial proxy model according to the strategy recommendation set to obtain a target proxy model; the target proxy model includes multiple base models and a meta-model that performs prediction based on the outputs of the multiple base models;

[0012] Deducing each of the initial strategies through multiple basic models in the target proxy model to obtain a pseudo-calculation result;

[0013] Screening the at least two initial strategies according to the pseudo calculation result to obtain a second reference strategy;

[0014] Deducing each second reference strategy through the strategy calculation model to obtain a second benchmark deduction result, and updating the strategy recommendation set according to the second reference strategy and the second benchmark deduction result;

[0015] Acquire a control demand target input by a user, determine a target strategy from the strategy recommendation set according to the control demand target, and recommend the target strategy to the user.

[0016] Optionally, the candidate strategies in the strategy recommendation set are used to perform multiple city optimization tasks;

[0017] Determining a target strategy from the strategy recommendation set according to the regulation demand target includes:

[0018] If the strategic goal is one of the multiple city optimization tasks, setting a first weight for the city optimization task corresponding to the strategic goal, and setting a second weight for the other city optimization tasks, wherein the first weight is greater than the second weight;

[0019] Scoring multiple candidate strategies in the strategy recommendation set according to the first weight and the second weight to obtain a strategy recommendation score for each candidate strategy;

[0020] Sort the candidate strategies from high to low according to the strategy recommendation scores to obtain a strategy ranking result;

[0021] According to the strategy ranking result, the candidate strategy corresponding to the highest strategy recommendation score is determined as the target strategy.

[0022] Optionally, after determining the candidate strategy corresponding to the strategy recommendation score with the highest score as the target strategy according to the strategy ranking result, the method further includes:

[0023] Traversing the candidate strategies in the strategy ranking result in turn, calculating the distance between the target strategy and each candidate strategy to obtain a strategy distance;

[0024] If the strategy distance is greater than a predetermined threshold, determining the candidate strategy as an auxiliary strategy;

[0025] The auxiliary strategy is recommended to the user.

[0026] Optionally, screening the at least two initial strategies according to the pseudo calculation result to obtain a second reference strategy includes:

[0027] Determining, based on the pseudo-calculation results, an expected improvement value resulting from executing each of a plurality of urban optimization tasks using the initial strategy; wherein the plurality of urban optimization tasks include at least one of the following: an implementation cost reduction task, an environmental impact reduction task, and an implementation time reduction task;

[0028] Obtaining an optimization weight set for each of the city optimization tasks;

[0029] Performing a weighted sum based on the optimization weight and the expected improvement value to obtain a target optimization score;

[0030] The at least two initial strategies are selected from high to low according to the target optimization score to obtain the second reference strategy.

[0031] Optionally, determining the expected improvement value of each of the multiple city optimization tasks performed by the initial strategy according to the pseudo calculation result includes:

[0032] Calculating a pseudo-calculation mean and a pseudo-calculation standard deviation based on the pseudo-calculation results output by the multiple basic models;

[0033] Performing difference calculation based on the current optimal function value, the pseudo-calculation mean, and a preset balance parameter to obtain a pseudo-calculation weight;

[0034] Performing ratio calculation based on the pseudo-calculation weight and the pseudo-calculation standard deviation to obtain a pseudo-calculation ratio;

[0035] Mapping the pseudo-calculation ratio through a preset cumulative distribution function to obtain a pseudo-calculation distribution function;

[0036] Multiplying the pseudo calculation weight and the pseudo calculation distribution function to obtain a first function term;

[0037] Mapping the pseudo-calculation ratio through a preset probability density function to obtain a pseudo-calculation probability density function;

[0038] Obtain a second function term by multiplying the pseudo-calculated standard deviation and the pseudo-calculated probability density function;

[0039] Summing the first function term and the second function term to obtain an expected improved function;

[0040] The expected improvement value is calculated according to the expected improvement function.

[0041] Optionally, the training of the initial proxy model according to the policy recommendation set to obtain a target proxy model includes:

[0042] Dividing the strategy recommendation set according to the total number of multiple basic models of the initial proxy model to obtain a strategy recommendation subset for each basic model;

[0043] Training the basic model according to the strategy recommendation subset of each basic model to update the basic model;

[0044] Deducing the first reference strategy in the strategy recommendation subset using each of the basic models to obtain a basic prediction result;

[0045] The meta-model of the initial proxy model is trained according to the first reference strategies and the basic prediction results corresponding to the multiple basic models to obtain the target proxy model.

[0046] Optionally, sampling the at least two initial strategies to obtain at least two first reference strategies includes:

[0047] Layering the at least two initial strategies to obtain multiple strategy layers;

[0048] Randomly selecting the initial strategy in each strategy layer to obtain a first reference strategy for each strategy layer;

[0049] The first reference policies of the multiple policy layers are shuffled to obtain the at least two first reference policies.

[0050] To achieve the above objectives, a second aspect of an embodiment of the present application provides an artificial intelligence-driven urban system computing simulation platform device, the device comprising:

[0051] A data acquisition module is used to acquire basic urban data used for urban system calculations, perform simulation calculations on the basic urban data, and obtain urban calculation element data;

[0052] A strategy generation module, configured to generate at least two initial strategies based on the city calculation element data;

[0053] a strategy sampling module, configured to sample the at least two initial strategies to obtain at least two first reference strategies;

[0054] a first deduction module, configured to deduce each of the first reference strategies using a pre-trained strategy calculation model to obtain a first benchmark deduction result, and add the first reference strategy and the first benchmark deduction result to a strategy recommendation set;

[0055] An agent model training module is used to train the initial agent model according to the policy recommendation set to obtain a target agent model; the target agent model includes multiple base models and a meta-model that performs prediction based on the outputs of the multiple base models;

[0056] A second deduction module is configured to deduce each of the initial strategies using a plurality of basic models in the target proxy model to obtain a pseudo calculation result;

[0057] a strategy screening module, configured to screen the at least two initial strategies according to the pseudo calculation result to obtain a second reference strategy;

[0058] a third deduction module, configured to deduce each of the second reference strategies using the strategy calculation model to obtain a second benchmark deduction result, and update the strategy recommendation set based on the second reference strategies and the second benchmark deduction result;

[0059] The strategy recommendation module is used to obtain the control demand target input by the user, determine the target strategy from the strategy recommendation set according to the control demand target, and recommend the target strategy to the user.

[0060] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the computer program.

[0061] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0062] The AI-driven urban system computational simulation platform method, apparatus, device, and medium proposed in this application considers the uncertainty of urban system development and simulates urban computational element data (such as the scale, function, structure, spatial planning, and other factors of the urban system) to generate initial strategies. Furthermore, rather than directly evaluating all initial strategies and predicting the specific effects and dynamic changes of the urban system after implementation, all initial strategies are first sampled and then calculated using an urban computational model to obtain an initial set of recommended strategies. Furthermore, a target proxy model is trained based on the recommended strategy set. The strategies are then calculated using multiple base models within the target proxy model to obtain pseudo-calculation results. The most appropriate strategy is selected based on the pseudo-calculation results, inferred using the urban computational model, and added to the recommended strategy set. This process is performed asynchronously to continuously refine the recommended strategy set. Finally, the most appropriate strategy is selected based on the user-entered control requirements and the recommended strategy set, and recommended to the user. The recommended strategies are used to help users find appropriate strategies across multiple objective dimensions (implementation cost, environmental impact, and implementation time) to meet their control needs.

[0063] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is an architecture diagram of the system used by the artificial intelligence-driven urban system computing simulation platform method provided in the embodiments of the present application;

[0065] Figure 2 This is a flow chart of the artificial intelligence-driven urban system computing simulation platform method provided in an embodiment of the present application;

[0066] Figure 3 This is another flow chart of the artificial intelligence-driven urban system computing simulation platform method provided in an embodiment of the present application;

[0067] Figure 4 yes Figure 2 Flowchart of step 203;

[0068] Figure 5 yes Figure 2 Flowchart of step 207;

[0069] Figure 6 yes Figure 2 Flowchart of step 209;

[0070] Figure 7 It is a structural diagram of the artificial intelligence-driven urban system computing simulation platform device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0072] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0074] The artificial intelligence-driven urban system computing simulation platform method provided in the embodiments of the present application can be applied to any terminal or server, and can also be software running on the server or terminal. The server can be configured as an independent physical server, or as a server cluster or distributed system consisting of multiple physical servers. It can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application or computer program that implements the artificial intelligence-driven urban system computing simulation platform method, but is not limited to the above forms.

[0075] The artificial intelligence-driven urban system computing simulation platform method, artificial intelligence-driven urban system computing simulation platform device, computer equipment, and computer-readable storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the artificial intelligence-driven urban system computing simulation platform method in the embodiments of the present application is described.

[0076] This application provides a method / technology for an AI-driven urban system computational simulation platform. This platform integrates computation, simulation, evaluation, visualization, and data analysis capabilities into a digital spatial planning and urban planning decision-making support platform. It is primarily applicable to key applications such as optimizing spatial development patterns and improving integrated transportation systems.

[0077] Reference Figure 1,This application is based on artificial intelligence technologies such as SMBO, DeepSeek and RAG, and realizes three major contents: data management system, urban computing engine and customer service platform.

[0078] 1. Data management system, mainly used for data acquisition and management:

[0079] The data management system is a process-based tool for transferring data from local storage to server (database) storage. It provides two main functions: data standardization and verification, and automated parameter calculation. It implements data processing functions such as automated review, import, cleansing and calculation, and reclassification and upload of initial complex data, providing data to the city calculation engine. The data management platform includes two core functions. The first is the "import" function, which mainly performs automated standardized review of various types of data in the "city standard information library" to ensure that the data items are complete and the data attribute fields meet the requirements before they can be imported into the city calculation engine. The second is the "calculation" function, which mainly implements the automated calculation of regional parameters. This function is mainly aimed at urban systems of different types and sizes. Through automated calculation, a set of regional parameters for urban systems is obtained, thereby ensuring the universal applicability of the CitySPS platform when operating in different urban systems.

[0080] (1) The “Import” function specifically includes: 1. Standardization review: Review the folder name, file name, file type, file encoding, unique identifier, number of data rows, number of data columns, and the format of each column of data before storage. 2. Completeness review: Review whether the data quantity meets the calculation requirements. This function requires the parameter automation function as a prerequisite. 3. Data cleaning and import into the database: If data does not meet the calculation conditions, the review result list will be returned. If it can be automatically corrected, it will be automatically cleaned and corrected. If it cannot be automatically corrected, it will be returned to the data collector for cleaning and modification. After the conditions are met, it will be imported into the production database.

[0081] (2) The "Calculation" function includes the automatic calculation of regional parameters, specifically the calculation of parameters required for the subsequent formal "Urban Development Trend Deduction" calculation, completing the process from data collection, parameter calculation to formal calculation. The specific calculation process includes: 1. Land module, calculating the number of buses in the grid, the growth of the number of buses, the number of subway lines, etc. 2. Population module, calculating the employment-residence coefficient, pre-processing grid population data, etc. 3. Transportation module, generating bus data, calculating data such as car ownership rate. The core calculation principles include the use of crawlers, linear regression, etc. for parameter calculation and data acquisition.

[0082] 2. City Calculation Engine:

[0083] The Urban Computing Engine is the platform's simulation computing support module and the code implementation of its core algorithm, specifically the principle algorithm for the city-wide system measurement model. Powered by a high-performance server hardware and software environment, it executes core urban computing functions, supporting the platform's core computations, such as urban computing and decision-making simulation, and providing users with analytical and computational data. The Urban Computing Engine consists of four internal modules: the data import layer, the data persistence and management layer, the technical solution layer, and the function and access layer.

[0084] The function of the data import layer is to receive the standardized variables and regional parameters of the data management platform. It quantifies the data collected in the early stage including the city's population, land, transportation, industry and other fields, and outputs the standard data format required for system calculation. At the same time, in view of the current situation that the data of each city has different regional characteristics, it is necessary to conduct multi-model training on the historical data of the corresponding city to obtain the regional parameters unique to each city. Finally, the parameters and the standardized data are uniformly imported into the urban system calculation simulation platform to start the system prediction calculation.

[0085] The data persistence and management layer systematically stores and manages various data items using database technology. This platform's database contains data with diverse attributes across multiple cities. Therefore, a storage approach employs a schema for each city to isolate and secure data, preventing data confusion and operational errors between cities and facilitating data management and maintenance. Furthermore, within a single city schema, the platform maintains a user table to store all user IDs. Furthermore, each user ID corresponds to multiple computing projects created by that user, resulting in project tables, as shown in Tables 1 and 2.

[0086] Table 1

[0087]

[0088] Table 2

[0089]

[0090] The technical solution layer encompasses big data technologies, quantitative simulation technologies, urban element deduction technologies, and scenario analysis technologies, supporting platform computations. This layer also serves as the core functional layer supporting the urban computing engine's computational work. First, big data technologies are used to process the underlying input data required for computations. These include basic geographic information (such as administrative division maps, road network vector data, and land use data), multi-source spatiotemporal big data (such as mobile phone signaling data and point of interest data), statistical data (such as multi-year census data and city statistical yearbooks), and resident travel survey data. Quantitative simulation technology provides computational support for completing "urban decision-making scenario simulations," specifically through systematic quantitative analysis and simulation of factors within a city, such as land, population, ecology, industrial chain, and transportation. Urban element deduction technology, based on the core algorithm "AI-enabled urban system model construction and simulation technology," provides computational support for simulating urban development trends. Scenario analysis technology supports urban decision-making recommendations within "urban decision-making scenario simulations," simulating and predicting urban population and employment distribution, traffic demand distribution, and transportation mode and route allocation.

[0091] The Function and Access Layer directly provides the customer service platform with user application functions such as calculations, data result browsing, and data access support. This layer is reflected in the platform's overall interface, including modules such as urban development trend simulation and urban decision-making scenario simulation. The Function and Access Layer comprises "Functional Modules" and "Access Modules." First, the "Functional Modules," along with the Analysis and Calculation Module, the Data Management Module, and the System Management Module, primarily provide functional services to users. The Analysis and Calculation Module includes the urban calculation engine, which performs a complete computation. This module sequentially runs the Urban Land Use Simulation and Evolution Module, the Population and Employment Distribution Module, the Traffic Demand Distribution Module, and the Transportation Mode Sharing and Route Allocation Module for computational analysis. Users obtain calculation results from the Analysis and Calculation Module, which are then transmitted to the Data Management Module as output indicators. The Data Management Module standardizes the received output indicator data and provides users with data storage and search capabilities based on computational projects. Finally, the System Management Module provides comprehensive account management, including backend account management, software version management, and a help center. The Access Module is a supporting module that ensures the normal operation of functional services and includes an API gateway, reverse proxy, and load balancing. The API gateway function is an intermediate layer located between the client and the backend. It serves as an entry point to process multiple API calls from the client, including access authentication, user authentication, access control and other functions. Users can access the platform through the API gateway function; the reverse proxy is a server. The user sends a request to the reverse proxy server through the API gateway. The reverse proxy server selects the target server to obtain data and then returns the data to the client. The use of the reverse proxy avoids direct interaction between the client and independent microservices, reduces the number of service calls, and controls traffic; the load balancing collects the call information of the backend service, such as the number of connections, response time, etc., and horizontally expands the cluster based on the collected information, distributes the requests to multiple processing nodes, and shares the processing pressure of the system, thereby improving the system operation efficiency and performance.

[0092] This platform utilizes a converged architecture for functional development. The platform's engineering architecture is primarily based on a browser / server model (B / S architecture), utilizing a front-end / back-end separation model. The front-end is developed using the VUE framework. The back-end server provides basic services such as system management, service scheduling, and data access, responding to user requests and returning data. Interaction with the front-end is achieved through a RESTful API, primarily using the lightweight JSON data exchange format for data exchange.

[0093] The platform's backend service hardware consists of four types: interface servers, computing servers, database servers, and scheduling servers. The computing servers are the core components of the backend module, primarily responsible for deploying and running the algorithmic model code of the CitySPS platform model. They perform analytical calculations on high-performance computing servers and push the results to the database servers. The server hardware utilizes a multi-core, high-performance CPU with a high capacity of RAM. The model algorithm code runs in multiple processes to improve computational efficiency.

[0094] 3. Customer Service Platform:

[0095] The customer service platform is an external service module and user operation platform. It runs on a browser and enables user interaction, displays city calculation results, and provides visualization. It also provides users with core functions for intelligent urban decision-making: ① urban development trend deduction, ② urban decision-making scenario simulation, and ③ LLM artificial intelligence semantic assistance system.

[0096] It should be noted that the artificial intelligence-driven urban system computing simulation platform method provided in this application is mainly based on the urban decision-making scenario simulation function.

[0097] ① Urban development trend deduction:

[0098] The urban development trend deduction function integrates spatial status information resources based on multi-dimensional big data. By constructing an urban system model and applying urban mechanism algorithms or urban machine learning algorithms, it simulates and evaluates the evolution trends of the core elements of the urban system during the planning cycle from the perspectives of total population and spatiotemporal distribution, land use scale and land function evolution, urban traffic routes and personal travel chains, and the matching relationship between supply and demand of service facilities, providing a reference for urban system planning.

[0099] ②Urban decision-making scenario simulation:

[0100] The urban decision-making scenario simulation function focuses on spatial planning and development, simulating multiple factors such as the scale, function, structure, and spatial planning of urban systems. This module simulates the implementation of over 20 strategies, identifying and comparing typical characteristics before and after implementation, providing a reference for urban construction and management. An intelligent decision support system is constructed, utilizing reinforcement learning algorithms to simulate and optimize various strategies for urban systems. For example, in urban planning, different land use strategies are simulated to assess their impact on transportation and the environment, and the optimal strategy is selected. A decision-making evaluation mechanism is established to evaluate the effectiveness of implemented strategies and provide feedback. Deep learning algorithms are used to analyze the results of implemented strategies, assess their impact on urban systems, and provide a reference for subsequent decision-making.

[0101] ③Semantic help system:

[0102] The LLM (Large Language Model) language assistance system, based on DeepSeek and RAG, first classifies the user's question to determine whether it is relevant to the document. If the question is relevant, RAG retrieves the relevant context, then generates an initial answer, further refines the answer, and ultimately returns it to the user. If the question is unrelated, sentiment analysis is performed, followed by a polite response generated and returned to the user.

[0103] To optimize the user experience, we use the DeepSeek model to categorize user questions into document-related and document-irrelevant. This categorization, based on pre-set prompts, determines the subsequent response strategy: document-related questions are answered through knowledge base retrieval, while document-irrelevant questions receive general responses.

[0104] For document-related questions, RAG (Retrieval Augmented Generation) technology is used. First, the document is loaded using TextLoader, and the text is segmented into searchable segments using TextSplitterFactory. Subsequently, FAISS is used to vectorize the text segments and store the results for fast retrieval. During the retrieval phase, a combination of keyword matching and vector retrieval is used to extract documents relevant to the question from the document repository. Specifically, Jieba word segmentation is used to extract keywords from the question. Two types of searches are then performed in parallel: a keyword-matching search, which scores the frequency of keywords in the document; and a vector-similarity search, which retrieves relevant documents by calculating the vector similarity between the question and the document. The results of these two searches are combined and sorted based on weights (for example, a keyword weight of 0.6 and a vector weight of 0.4). After deduplication, the top five documents are selected to form the context. The entire retrieval process is executed asynchronously to improve efficiency, and hybrid ranking is used to optimize result quality, ultimately generating the most relevant contextual content. After the search is complete, the DeepSeek model is used to generate an initial answer, combining the retrieved context with the user's question, ensuring that the answer is concise and fully utilizes contextual information. If no relevant context is found, the user is informed that the answer cannot be given. Finally, the DeepSeek model is used to optimize the answer, removing unnecessary or duplicate content and retaining the core information to obtain the final answer.

[0105] For questions unrelated to the document, we first use the DeepSeek model to analyze the user's sentiment, including emotional inclination, urgency, and whether they need reassurance. Based on the results of the sentiment analysis, we then use the DeepSeek model again to generate a friendly and professional response that meets the user's emotional needs.

[0106] The LLM semantic help system developed on this platform accurately implements precise semantic understanding and generation functions based on existing artificial intelligence technologies, provides users with efficient and intelligent semantic help services, meets diverse platform semantic processing needs, and realizes the application of artificial intelligence in urban complex system calculation and simulation platforms.

[0107] Please refer to Figure 2 , Figure 2 An optional flowchart of an artificial intelligence-driven urban system computational simulation platform method is disclosed. Figure 2 The method may include but is not limited to steps 201 to 209.

[0108] Step 201: Obtain city basic data for city system calculation, perform simulation calculation on the city basic data, and obtain city calculation element data;

[0109] Step 202: Generate at least two initial strategies based on the city calculation element data;

[0110] Step 203: sampling at least two initial strategies to obtain at least two first reference strategies;

[0111] Step 204: deduce each first reference strategy using the pre-trained strategy calculation model to obtain a first benchmark deduction result, and add the first reference strategy and the first benchmark deduction result to the strategy recommendation set;

[0112] Step 205: Train the initial proxy model according to the policy recommendation set to obtain a target proxy model; the target proxy model includes multiple base models and a meta-model that performs prediction based on the outputs of the multiple base models;

[0113] Step 206, deduce each initial strategy using multiple basic models in the target proxy model to obtain a pseudo calculation result;

[0114] Step 207: Screen at least two initial strategies based on the pseudo calculation result to obtain a second reference strategy;

[0115] Step 208: deduce each second reference strategy using the strategy calculation model to obtain a second benchmark deduction result, and update the strategy recommendation set based on the second reference strategy and the second benchmark deduction result;

[0116] Step 209 : obtaining the control demand target input by the user, determining a target strategy from the strategy recommendation set according to the control demand target, and recommending the target strategy to the user.

[0117] In steps 201 to 209, as illustrated in this embodiment of the present application, simulations are performed on urban computing element data (such as the scale, function, structure, spatial planning, and other factors of the urban system) to generate initial strategies. Furthermore, rather than directly evaluating all initial strategies and predicting the specific effects and dynamic changes in the urban system after implementation, all initial strategies are first sampled and then calculated using the urban computing model to generate an initial set of recommended strategies. Furthermore, a target proxy model is trained based on the recommended strategy set. The strategies are then calculated using multiple base models within the target proxy model to generate pseudo-calculation results. The most appropriate strategy is selected based on the pseudo-calculation results, inferred using the urban computing model, and added to the recommended strategy set. This process is performed asynchronously, continuously refining the recommended strategy set. Finally, the most appropriate strategy is selected based on the user-entered control requirements and the recommended strategy set, and recommended to the user. Strategy recommendations are used to help users find appropriate strategies across multiple objective dimensions (implementation cost, environmental impact, and implementation time) to meet their needs.

[0118] In one example, referring to Figure 3 , AI-driven urban system computational simulation platform methods can include:

[0119] (1) Generate all strategy combinations: Generate at least two initial strategies;

[0120] (2) Strategy sampling: Sampling at least two initial strategies to obtain at least two first reference strategies;

[0121] (3) Urban computational model deduction strategy: Deducing each first reference strategy through the urban computational model to obtain the first benchmark deduction result;

[0122] (4) Update the strategy recommendation set: add the first reference strategy and the first benchmark deduction result to the strategy recommendation set;

[0123] (5) Proxy model training: The initial proxy model is trained according to the policy recommendation set to obtain the target proxy model;

[0124] (6) Candidate strategy evaluation: Use the trained target proxy model to predict each initial strategy and obtain an approximate calculation result, and then filter the initial strategies based on the approximate calculation result;

[0125] (7) Prediction and control results of the proxy model: Each initial strategy is deduced through multiple basic models in the target proxy model to obtain pseudo-calculation results;

[0126] (8) Select the optimal strategy: Screen at least two initial strategies based on the pseudo-calculation results to obtain a second reference strategy;

[0127] (9) Urban calculation model deduction strategy: Deducing each second reference strategy through the strategy calculation model to obtain the second benchmark deduction result, and then entering (4) to update the strategy recommendation set based on the second reference strategy and the second benchmark deduction result;

[0128] (10) User sets control demand target: obtains the control demand target input by the user;

[0129] (11) Obtaining strategy recommendation results: Determine the target strategy from the strategy recommendation set based on the control demand target;

[0130] (12) Make strategy recommendations: recommend target strategies to users.

[0131] In step 201, basic urban data includes basic geographic information data (such as administrative division map data, road network vector data, and land use data), multi-source spatiotemporal big data (such as mobile phone signaling data and point of interest data), statistical data (such as multi-year census data, economic census data, and city statistical yearbooks), and resident travel survey data. Basic urban data may also include basic urban image data (such as census data charts and city statistical annual inspection charts), basic urban text data (such as administrative division map data, road network vector data, and land use data), and other data.

[0132] Quantitative simulation and urban element deduction techniques can be used to simulate and deduce basic urban data to obtain urban computational element data. Quantitative simulation technology provides the computational support for completing a "city decision-making scenario simulation." Specifically, it involves systematically analyzing and simulating factors within a city, such as land, population, ecology, industrial chain, and transportation, to obtain urban computational element data. For a detailed description of quantitative simulation technology, please refer to the previous article and will not be elaborated here.

[0133] In step 202, at least two initial strategies are generated based on the urban calculation element data. An initial strategy includes at least one strategy, i.e., it is a strategy combination. For example, an initial strategy may include a strategy for "increasing the proportion of green buildings." Another example may include a strategy for "adjusting the land supply structure." Another example may include a strategy for "increasing the proportion of green buildings" and a strategy for "adjusting the land supply structure."

[0134] In one example, in urban planning, multiple strategies are generated based on urban computational element data. At least two of these strategies are permuted and combined to produce a variety of different land use strategies. This can be generated using a large language model or other methods, which are not specifically limited in this application.

[0135] In step 203, at least two initial strategies are sampled to obtain at least two first reference strategies. The sampling method may be random sampling.

[0136] In one embodiment, referring to Figure 4 , step 203 may include:

[0137] Step 401, layering at least two initial strategies to obtain multiple strategy layers;

[0138] Step 402: Randomly select the initial strategy in each strategy layer to obtain the first reference strategy of each strategy layer;

[0139] Step 403: Randomize the first reference policies of multiple policy layers to obtain at least two first reference policies.

[0140] In step 401, different initial policies have different numbers of policies. The number of policies in at least two initial policies can be stratified to obtain multiple policy layers. For example, the number of policies in the first policy layer can be 1, the number of policies in the second policy layer can be 2, the number of policies in the third policy layer can be 3, and so on. In this way, stratification based on the number of policies can closely reflect the characteristic of the initial policy as a combination of policies, greatly improving the accuracy of the stratification.

[0141] In step 402, at least one first reference strategy may be randomly selected from at least one initial strategy in each strategy layer. For example, two first reference strategies may be randomly selected from the first strategy layer, two first reference strategies may be randomly selected from the second strategy layer, and one first reference strategy may be randomly selected from the third strategy layer.

[0142] In one embodiment, step 402 may include: determining a selection threshold based on the number of policies corresponding to each policy layer; randomly selecting the initial policies in each policy layer based on the selection area to obtain a first reference policy with a number of policies less than or equal to the selection threshold. The relationship between the number of policies and the selection threshold is an inverse relationship, that is, the larger the number of policies, the smaller the selection threshold. For example, if the number of policies in the first policy layer is 1, the selection threshold is 10; if the number of policies in the second policy layer is 2, the selection threshold is 5; if the number of policies in the third policy layer is 3, the selection threshold is 3. In this way, determining the selection threshold for each policy layer based on the number of policies can effectively reduce the number of policies in the selected first reference policy, increase the probability that the first reference policy passes the subsequent evaluation, and improve the reliability and accuracy of policy selection.

[0143] In step 403, the selected first reference strategies are shuffled to improve the randomness of subsequent strategy evaluations, thereby improving evaluation accuracy.

[0144] The benefit of the embodiment of the above steps 401 to 403 is that the rationality and randomness of the distribution of the strategies selected during the sampling process are improved, thereby improving the sampling accuracy.

[0145] In step 204, each first reference strategy is deduced using the pre-trained strategy calculation model to obtain a first benchmark deduction result. The first reference strategy and the first benchmark deduction result are then added to the strategy recommendation set. The strategy calculation model is a pre-trained machine learning model used for strategy deduction. It can deduce the execution effect / performance level of multiple city optimization tasks based on the input strategy. However, the model has many parameters and the deduction speed is relatively slow.

[0146] Specifically, since the initial strategy is a combination of strategies, different initial strategies are actually combinations of different strategies. The large number of strategies and the fact that they can be combined arbitrarily result in a large number of initial strategies. Existing techniques often evaluate all initial strategies separately to select the optimal one, but this evaluation is extremely inefficient and difficult to generalize. In this application, sampling is performed first, and then the strategy calculation model is used to deduce each sampled first reference strategy, reducing the number of strategies that need to be deduced and thereby improving deduction efficiency.

[0147] The multiple urban optimization tasks include at least one of the following: reducing implementation costs, reducing environmental impacts, and shortening implementation time. For example, an initial strategy, "Increasing the proportion of green buildings," is input into the strategy calculation model. The strategy calculation model then performs a deduction to obtain a first benchmark deduction result, which typically includes various indicators for the forecast year, the implementation cost of the strategy, and the implementation time. The first reference strategy is denoted as x1, the first benchmark deduction result is denoted as f(x1), and the recommended strategy set is denoted as H. After adding x1 and f(x1) to H, H = {(x1, f(x1))} is obtained.

[0148] In step 205, the initial proxy model is trained based on the policy recommendation set to obtain a target proxy model. The target proxy model includes multiple base models and a meta-model that performs predictions based on the outputs of these base models. The target proxy model has fewer model parameters than the city calculation model, resulting in a relatively high deduction speed.

[0149] In one embodiment, step 205 may include: predicting each first reference strategy through each basic model to obtain an initial pseudo-calculation result output by each basic model; predicting the initial pseudo-calculation results output by multiple basic models through a meta-model to obtain an initial pseudo-deduction result; performing loss calculation based on the initial pseudo-deduction result and the first benchmark deduction result to obtain a proxy loss function; and adjusting parameters of multiple basic models and meta-models based on the proxy loss function to obtain a target proxy model.

[0150] The above loss calculation can use the mean square error loss function or other loss functions, which is not specifically limited in this application.

[0151] In one example, an initial proxy model is trained based on a set of recommended policies H to obtain a target proxy model S. The target proxy model S can efficiently learn the mapping and relationship between input policies and output outcomes. To improve the effectiveness of the target proxy model, an ensemble learning approach is used to construct the proxy model, leveraging the strengths of multiple machine learning algorithms to improve overall prediction accuracy. Basic ensemble learning models include Random Forest Regression, Gradient Boosting, Multilayer Perceptron (MLP), XGBoost, and Gaussian Process Regression. The meta-model uses Ridge Regression, which is trained on the prediction results of the basic models.

[0152] In another embodiment, step 205 may include: dividing the policy recommendation set according to the total number of multiple basic models of the initial proxy model to obtain a policy recommendation subset for each basic model; training the basic model according to the policy recommendation subset of each basic model to update the basic model; deducing the first reference policy in the policy recommendation subset through each basic model to obtain a basic prediction result; training the meta-model of the initial proxy model according to the first reference policy and basic prediction results corresponding to the multiple basic models to obtain a target proxy model. The main difference between this embodiment and the previous embodiment is that a policy recommendation subset is divided from the policy recommendation set for each basic model, and the policy recommendation subsets between different basic models do not affect each other, which not only reduces the complexity of model training to improve model training efficiency, but also improves the deduction accuracy.

[0153] In one example, all possible strategies are combined to obtain the initial strategy, and then the trained proxy model S is used to predict these combinations to obtain the target calculation result S(x). , , and metamodel , input strategy x, each base model generates pseudo calculation results The meta-model generates target calculation results based on these pseudo calculation results .

[0154] Before step 206 , the initial strategies may be preliminarily screened based on the target calculation result S(x) to further reduce the number of strategies and improve processing efficiency.

[0155] In step 206, each initial strategy is deduced through multiple basic models in the target proxy model to obtain a pseudo calculation result. For example, referring to the above, for K basic models , , input strategy x, each base model generates pseudo calculation results .

[0156] In step 207, at least two initial strategies are screened according to the pseudo calculation result to obtain a second reference strategy.

[0157] In one embodiment, referring to Figure 5 , step 207 may include:

[0158] Step 501, determining the expected improvement value generated by executing each of the multiple city optimization tasks using the initial strategy based on the pseudo-calculation results;

[0159] Step 502: Obtain the optimization weight set for each city optimization task;

[0160] Step 503: Perform a weighted sum based on the optimization weight and the expected improvement value to obtain a target optimization score;

[0161] Step 504 : Select at least two initial strategies from high to low according to the target optimization scores to obtain a second reference strategy.

[0162] In one embodiment, step 501 may include:

[0163] According to the pseudo-calculation results output by multiple basic models, the pseudo-calculation mean and pseudo-calculation standard deviation are calculated;

[0164] Perform difference calculation based on the current optimal function value, pseudo-calculation mean, and preset balance parameters to obtain pseudo-calculation weights;

[0165] Perform ratio calculation based on pseudo-calculation weight and pseudo-calculation standard deviation to obtain pseudo-calculation ratio;

[0166] The pseudo-calculation ratio is mapped by a preset cumulative distribution function to obtain a pseudo-calculation distribution function;

[0167] The first function term is obtained by multiplying the pseudo-calculation weight and the pseudo-calculation distribution function;

[0168] The pseudo-calculation ratio is mapped by a preset probability density function to obtain a pseudo-calculation probability density function;

[0169] The second function term is obtained by multiplying the pseudo-calculated standard deviation and the pseudo-calculated probability density function;

[0170] Summing the first function term and the second function term to obtain an expected improved function;

[0171] According to the expected improvement function, the expected improvement value is calculated.

[0172] Specifically, the most appropriate strategy is selected from the calculation results based on the multi-objective weighted expected improvement (EI) and the urban calculation model is used to Reasoning about the strategy and obtaining accurate calculation results .Will Add to the strategy recommendation set H. Repeat this process to enrich the recommendation set. The expected improvement is used to select the acquisition function of the next evaluation point. For the minimization problem, the basic expected improvement formula is ,in is the currently observed best function value, and are the pseudo-calculated mean and pseudo-calculated standard deviation of the predictions obtained by each base model for strategy x. is the exploration-exploitation balance parameter, usually set to 0.01. , is the cumulative distribution function (CDF) of the standard normal distribution, is the probability density function (PDF) of the standard normal distribution. Since this application has multiple optimization tasks, multi-task optimization is required and the weighted method is used to calculate: .in: is the target optimization score of the initial policy x, is the weight of task j, which is user-defined. is the expected improvement value of target j. n is the number of tasks. In this application, tasks may include tasks for reducing implementation costs, reducing environmental impacts, and shortening implementation time.

[0173] The benefit of the embodiment of steps 501 to 504 is that it can perform multi-task evaluation on the initial strategy, improve screening accuracy, and has high applicability.

[0174] In step 208, each second reference strategy is deduced using the strategy calculation model to obtain a second benchmark deduction result. The strategy recommendation set is then updated based on the second reference strategy and the second benchmark deduction result. For example, let the second reference strategy be x2, the second benchmark deduction result be f(x2), and referring to the above, the strategy recommendation set H = {(x1, f(x1))}. After adding x2 and f(x2) to H, H = {(x1, f(x1)), (x2, f(x2))} is obtained. The first reference strategy and the second reference strategy in the strategy recommendation set are collectively referred to as candidate strategies. The candidate strategies in the strategy recommendation set are used to perform multiple urban optimization tasks.

[0175] In step 209 , the control demand target input by the user is obtained, a target strategy is determined from the strategy recommendation set according to the control demand target, and the target strategy is recommended to the user.

[0176] The control demand objective refers to the urban optimization task that the user expects the strategy to achieve. For example, a user might enter the control demand objective as "expected control implementation costs to be within 100 billion yuan." Another example might be "expected implementation time to be less than 5 years." Another example might be "expected control implementation costs to be within 100 billion yuan and expected implementation time to be less than 5 years."

[0177] In one embodiment, referring to Figure 6 The step 209 of determining the target strategy from the strategy recommendation set according to the control demand target may include:

[0178] Step 601: If the regulation demand target is one of the multiple city optimization tasks, a first weight is set for the city optimization task corresponding to the regulation demand target, and a second weight is set for the other city optimization tasks, wherein the first weight is greater than the second weight.

[0179] Step 602: Score multiple candidate strategies in the strategy recommendation set according to the first weight and the second weight to obtain a strategy recommendation score for each candidate strategy;

[0180] Step 603: sort the candidate strategies from high to low according to the strategy recommendation scores to obtain a strategy sorting result;

[0181] Step 604: Determine the candidate policy corresponding to the highest policy recommendation score as the target policy based on the policy ranking result.

[0182] For example, when a user enters a specific control requirement, the application filters the strategy recommendation set H based on the user's input to find strategies and their deduction results that meet the user's requirements. Finally, the target strategy and its deduction results are recommended to the user to meet the user's needs.

[0183] This application performs maximum and minimum normalization processing on each task, that is, , after standardization, calculate the weighted score Here The preference weights used when making recommendations for users. For example, if a user wants to see strategies with lower implementation costs, the implementation cost reduction task will have a higher weight, such as 0.8, while the weights of other tasks will be 0.1. Strategies with lower implementation costs will receive higher scores, and other costs will also affect the scores.

[0184] In one embodiment, in order to ensure the diversity of recommended strategies, the present application adds a diversity guarantee mechanism. Then, after step 604, an artificial intelligence-driven urban system computing simulation platform method may further include: traversing the candidate strategies in the strategy ranking results in sequence, performing distance calculations on the target strategy and each candidate strategy to obtain a strategy distance; if the strategy distance is greater than a predetermined threshold, determining the candidate strategy as an auxiliary strategy; and recommending the auxiliary strategy to the user.

[0185] For example, we first sort candidate strategies by score, select the strategy with the highest score as the target strategy, and begin adding it to the recommended results. We then iterate through each candidate strategy, calculating the normalized Euclidean distance between the candidate strategy and the selected strategy. If the distance is greater than a threshold, the strategy is added to the recommended results. This process is repeated until the required number of strategies is reached. This ensures that the recommended strategies are diverse, making it easier for users to select the strategy that best suits them.

[0186] In one example, a user inputs a regulatory objective: optimizing the urban system's land structure while simultaneously desiring to keep implementation costs below 100 billion yuan and implementation time less than five years. Using the AI-driven urban system computational simulation platform method provided in this application, recommendations are generated across three broad categories: "lowest implementation cost strategies" that control costs, "lowest environmental impact strategies" that manage environmental impacts, and "shortest implementation time strategies" that manage implementation time. Based on the regulatory objective, the three strategies under the "lowest implementation cost" category are primarily compared. Strategy 1, by increasing the proportion of green buildings, will incur implementation costs of 86.8 to 106.1 billion yuan, with an estimated implementation period of 36 months. Strategy 2, by adjusting the land supply structure, will incur implementation costs of 2,090.234 to 2,554.7305 billion yuan. Strategy 3, through comprehensive regulation (simultaneously increasing the proportion of green buildings and adjusting the land supply structure), will incur implementation costs of 2,509.3576 billion yuan to 3,066.9926 billion yuan. The recommended strategy selected this time is Strategy 1, which includes increasing the proportion of green buildings, from the lowest implementation cost strategy.

[0187] The various technical features in the above embodiments can be combined arbitrarily as long as there is no conflict or contradiction between the combinations of features. However, due to space limitations, they are not described one by one. Therefore, the arbitrary combination of the various technical features in the above embodiments also falls within the scope of disclosure of this specification.

[0188] The present application also discloses an artificial intelligence driven urban system computing simulation platform device, referring to Figure 7 The artificial intelligence-driven urban system calculation simulation platform device includes: a data acquisition module 701, which is used to obtain urban basic text data for urban system calculation, perform simulation calculation on the urban basic text data, and obtain urban calculation element data; a strategy generation module 702, which is used to generate at least two initial strategies according to the urban calculation element data; a strategy sampling module 703, which is used to sample at least two initial strategies and obtain at least two first reference strategies; a first deduction module 704, which is used to deduce each first reference strategy through a pre-trained strategy calculation model to obtain a first benchmark deduction result, and add the first reference strategy and the first benchmark deduction result to a strategy recommendation set; an agent model training module 705, which is used to train the initial agent model according to the strategy recommendation set to obtain to the target proxy model; the target proxy model includes multiple basic models and a meta-model for making predictions based on the outputs of the multiple basic models; a second deduction module 706 is used to deduce each initial strategy through the multiple basic models in the target proxy model to obtain a pseudo-calculation result; a strategy screening module 707 is used to screen at least two initial strategies according to the pseudo-calculation result to obtain a second reference strategy; a third deduction module 708 is used to deduce each second reference strategy through the strategy calculation model to obtain a second benchmark deduction result, and update the strategy recommendation set according to the second reference strategy and the second benchmark deduction result; a strategy recommendation module 709 is used to obtain the control demand target input by the user, determine the target strategy from the strategy recommendation set according to the control demand target, and recommend the target strategy to the user.

[0189] In one embodiment, the artificial intelligence-driven urban system computing simulation platform device also includes an auxiliary recommendation module, which is used to traverse the candidate strategies in the strategy ranking results in sequence, calculate the distance between the target strategy and each candidate strategy, and obtain the strategy distance; if the strategy distance is greater than a predetermined threshold, the candidate strategy is determined as an auxiliary strategy; and the auxiliary strategy is recommended to the user.

[0190] It should be noted that the specific implementation of the artificial intelligence-driven urban system computing simulation platform device is basically the same as the specific implementation of the above-mentioned artificial intelligence-driven urban system computing simulation platform method, and will not be repeated here.

[0191] The present application also discloses a computer device comprising a memory, a processor, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, the aforementioned AI-driven urban system computational simulation platform method is implemented. The computer device may be, for example, a mobile phone, a computer, or the like.

[0192] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned artificial intelligence-driven urban system computing simulation platform method.

[0193] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0194] An embodiment of the present application also provides a computer program product, which includes a computer program, which is read and executed by a processor, so that when the processor executes the computer program, it implements the above-mentioned artificial intelligence-driven urban system computing simulation platform method.

[0195] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0196] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0198] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0199] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0200] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0201] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0202] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0203] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0204] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, 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.

[0205] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. An artificial intelligence-driven urban system computational simulation platform method, characterized in that: The method comprises: Acquiring city basic data for city system calculation, performing simulation calculation on the city basic data, and obtaining city calculation element data; generating at least two initial strategies based on the city calculation element data; Sampling the at least two initial strategies to obtain at least two first reference strategies; Deducing each of the first reference strategies using a pre-trained strategy calculation model to obtain a first benchmark deduction result, and adding the first reference strategy and the first benchmark deduction result to a strategy recommendation set; Training the initial proxy model according to the strategy recommendation set to obtain a target proxy model; the target proxy model includes multiple base models and a meta-model that performs prediction based on the outputs of the multiple base models; Deducing each of the initial strategies through multiple basic models in the target proxy model to obtain a pseudo-calculation result; Screening the at least two initial strategies according to the pseudo calculation result to obtain a second reference strategy; Deducing each second reference strategy through the strategy calculation model to obtain a second benchmark deduction result, and updating the strategy recommendation set according to the second reference strategy and the second benchmark deduction result; Acquire a control demand target input by a user, determine a target strategy from the strategy recommendation set according to the control demand target, and recommend the target strategy to the user.

2. The method according to claim 1, characterized in that The candidate strategies in the strategy recommendation set are used to perform multiple city optimization tasks; Determining a target strategy from the strategy recommendation set according to the regulation demand target includes: If the regulation demand target is one of the multiple city optimization tasks, a first weight is set for the city optimization task corresponding to the regulation demand target, and a second weight is set for the other city optimization tasks, wherein the first weight is greater than the second weight; Scoring multiple candidate strategies in the strategy recommendation set according to the first weight and the second weight to obtain a strategy recommendation score for each candidate strategy; Sort the candidate strategies from high to low according to the strategy recommendation scores to obtain a strategy ranking result; According to the strategy ranking result, the candidate strategy corresponding to the highest strategy recommendation score is determined as the target strategy.

3. The method according to claim 2, characterized in that After determining the candidate strategy with the highest strategy recommendation score as the target strategy based on the strategy ranking result, the method further includes: Traversing the candidate strategies in the strategy ranking result in turn, calculating the distance between the target strategy and each candidate strategy to obtain a strategy distance; If the strategy distance is greater than a predetermined threshold, determining the candidate strategy as an auxiliary strategy; The auxiliary strategy is recommended to the user.

4. The method according to any one of claims 1 to 3, characterized in that The step of screening the at least two initial strategies according to the pseudo calculation result to obtain a second reference strategy includes: Determining, based on the pseudo-calculation results, an expected improvement value resulting from executing each of a plurality of urban optimization tasks using the initial strategy; wherein the plurality of urban optimization tasks include at least one of the following: an implementation cost reduction task, an environmental impact reduction task, and an implementation time reduction task; Obtaining an optimization weight set for each of the city optimization tasks; Performing a weighted sum based on the optimization weight and the expected improvement value to obtain a target optimization score; The at least two initial strategies are selected from high to low according to the target optimization score to obtain the second reference strategy.

5. The method according to claim 4, characterized in that Determining the expected improvement value generated by executing each of the multiple city optimization tasks according to the pseudo calculation results includes: Calculating a pseudo-calculation mean and a pseudo-calculation standard deviation based on the pseudo-calculation results output by the multiple basic models; Performing difference calculation based on the current optimal function value, the pseudo-calculation mean, and a preset balance parameter to obtain a pseudo-calculation weight; Performing ratio calculation based on the pseudo-calculation weight and the pseudo-calculation standard deviation to obtain a pseudo-calculation ratio; Mapping the pseudo-calculation ratio through a preset cumulative distribution function to obtain a pseudo-calculation distribution function; Multiplying the pseudo calculation weight and the pseudo calculation distribution function to obtain a first function term; Mapping the pseudo-calculation ratio through a preset probability density function to obtain a pseudo-calculation probability density function; Obtain a second function term by multiplying the pseudo-calculated standard deviation and the pseudo-calculated probability density function; Summing the first function term and the second function term to obtain an expected improved function; The expected improvement value is calculated according to the expected improvement function.

6. The method according to any one of claims 1 to 3, characterized in that The training of the initial proxy model according to the strategy recommendation set to obtain the target proxy model includes: Dividing the strategy recommendation set according to the total number of multiple basic models of the initial proxy model to obtain a strategy recommendation subset for each basic model; Training the basic model according to the strategy recommendation subset of each basic model to update the basic model; Deducing the first reference strategy in the strategy recommendation subset using each of the basic models to obtain a basic prediction result; The meta-model of the initial proxy model is trained according to the first reference strategies and the basic prediction results corresponding to the multiple basic models to obtain the target proxy model.

7. The method according to any one of claims 1 to 3, characterized in that The sampling of the at least two initial strategies to obtain at least two first reference strategies includes: Layering the at least two initial strategies to obtain multiple strategy layers; Randomly selecting the initial strategy in each strategy layer to obtain a first reference strategy for each strategy layer; The first reference policies of the multiple policy layers are shuffled to obtain the at least two first reference policies.

8. An artificial intelligence-driven urban system computing simulation platform device, characterized in that: The device comprises: A data acquisition module is used to acquire basic urban data used for urban system calculations, perform simulation calculations on the basic urban data, and obtain urban calculation element data; A strategy generation module, configured to generate at least two initial strategies based on the city calculation element data; a strategy sampling module, configured to sample the at least two initial strategies to obtain at least two first reference strategies; a first deduction module, configured to deduce each of the first reference strategies using a pre-trained strategy calculation model to obtain a first benchmark deduction result, and add the first reference strategy and the first benchmark deduction result to a strategy recommendation set; An agent model training module is used to train the initial agent model according to the policy recommendation set to obtain a target agent model; the target agent model includes multiple base models and a meta-model that performs prediction based on the outputs of the multiple base models; A second deduction module is configured to deduce each of the initial strategies using a plurality of basic models in the target proxy model to obtain a pseudo calculation result; a strategy screening module, configured to screen the at least two initial strategies according to the pseudo calculation result to obtain a second reference strategy; a third deduction module, configured to deduce each of the second reference strategies using the strategy calculation model to obtain a second benchmark deduction result, and update the strategy recommendation set based on the second reference strategies and the second benchmark deduction result; The strategy recommendation module is used to obtain the control demand target input by the user, determine the target strategy from the strategy recommendation set according to the control demand target, and recommend the target strategy to the user.

9. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which implements the method according to any one of claims 1 to 7 when executed.

Images