Multi-agent based full-time simulation optimization method for informal settlements in central business districts

Abstract

The application discloses a kind of center area non-regular living space full-time simulation optimization method based on multi-agent, including full-time non-regular living space space-time behavior information data acquisition and pre-processing, city center area non-regular living space function identification and index extraction, non-regular living space function-oriented differentiated weight evaluation system construction, non-regular living space inefficient area identification, full-time non-regular living space multi-agent model construction, non-regular living space optimization simulation based on multi-agent and non-regular living space optimization scheme visualization and intelligent control seven steps.The application realizes the intelligent perception and dynamic response of the behavior trajectory of people in non-regular living space in each period of the day, helps to promote the efficiency optimization and structure reasonable layout of non-regular living space in time dimension, improves the timeliness, accuracy and scientificity of planning scheme, and promotes the coordinated people's city construction in full-time.

Description

Technical Field

[0001] This invention belongs to the technical field of intelligent planning and design, specifically a method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems. Background Technology

[0002] With the deepening of urbanization, informal residential spaces are gradually emerging in urban centers, reflecting structural contradictions in urban governance and spatial resource allocation. Current research on these spaces largely focuses on static physical environment analysis, lacking a systematic exploration of the dynamic behavior of people within the region. This makes it difficult to reflect real-time changes in urban space, resulting in optimization measures lagging behind actual needs. Traditional research methods have limitations in three aspects: First, limited by data collection technology, traditional information acquisition mainly relies on manual surveys, questionnaires, and simple data analysis methods, which suffer from strong subjectivity, limited coverage, and insufficient accuracy, making it difficult to support the analysis needs of today's complex urban environments. Second, traditional evaluation systems are relatively singular, lacking multi-dimensional, full-scale comprehensive evaluation standards, making it difficult to fully reveal the complex characteristics of informal residential spaces. Finally, traditional methods are mostly based on historical data for empirical judgment and response strategy construction, failing to effectively integrate multi-source big data for future trend prediction, resulting in planning strategies lacking foresight and accuracy.

[0003] Multi-agent simulation, a technology used in the context of the new economy to simulate the impact of individual travel behaviors on urban space, is increasingly being applied to urban optimization and management. It can dynamically simulate the behavioral patterns of people in informal residential spaces in urban centers, achieving intelligent perception of people's spatiotemporal behavioral trajectories. This helps promote the dynamic simulation and efficiency optimization of informal spaces in urban centers throughout all time periods, improving the accuracy and scientific rigor of planning schemes. Summary of the Invention

[0004] The purpose of this invention is to address the problems of inefficiency and dynamic temporal changes in informal residential spaces that are neglected in traditional central area planning and design, and to provide a method for optimizing the real-time simulation of informal residential spaces in central areas based on multi-agent systems. This method enables intelligent perception and dynamic response to the behavioral trajectories of people in informal residential spaces throughout the day, which helps to optimize the efficiency and rationally arrange the structure of informal residential spaces in the time dimension, improve the timeliness, accuracy and scientific nature of planning schemes, and promote the construction of people-oriented cities that are coordinated throughout the day.

[0005] The technical solution adopted in this invention is: optimization of full-time simulation of irregular residential spaces in the central area based on multi-agent systems. The method includes the following steps: Step S1: Multi-source data collection and preprocessing of informal residential spaces throughout the entire time period. Within the boundary of a city center, multi-source heterogeneous data on informal spaces are collected. Spatial data includes city boundary data, basic land use data, building outline data, road network data, and geometric attribute data of the informal space itself. For population dynamics data, information on population dwelling, movement, and gathering behavior trajectories throughout any seven days is identified and collected using mobile phone signaling data, WiFi probe data, and video surveillance data. The all-time data is divided into four time periods: morning, noon, evening, and night. The acquired population dynamics data is preprocessed, with an anchoring time threshold of 10 minutes and a spatial threshold of 50 meters, removing abnormal and duplicate data. The datasets after the above filtering are stored in a distributed database and a geospatial database to construct a full-time spatiotemporal behavior database for informal residential spaces.

[0006] Step S2: Functional Identification and Indicator Extraction of Informal Residential Spaces in Urban Centers. Based on the database from Step S1, the study area is divided into spatial functional units and identified by type. The study area is divided into basic grids of 50m × 50m. For each grid unit, multidimensional feature data is extracted. Based on its dominant function, each evaluation unit is divided into living-dominated, production-dominated, or ecology-dominated spaces, and multivariate cluster analysis is performed. From the three dimensions of living, production, and ecology, a multi-level comprehensive evaluation index system for informal residential spaces is constructed, including 9 primary indicators and 16 secondary indicators. Finally, the index dataset is integrated to unify the different indicator dimensions. All data are linearly transformed to the range of 0 to 100% using a standardized formula to achieve data normalization.

[0007] Step S3: Construction of a differentiated weighted evaluation system for informal residential spaces under functional orientation. Based on the indicator dataset established in Step S2, perform classification calculations and comprehensive evaluations. Calculate the information entropy of each indicator. The specific formula is as follows If there are n indicators: j=1,2,...n, and i is the sample number, and if we are analyzing n grid regions, then i=1,2,...,n, where n is the total number of samples and m is the total number of grids. Let be the weight of the i-th sample on the j-th indicator; dynamically construct a differentiated weight set, and use the weight set corresponding to its spatial type for each evaluation unit, assigning higher weights to indicators closely related to the core functions, and calculate its comprehensive performance score to obtain the weight. Based on this, a comprehensive performance evaluation model for informal residential spaces is constructed, and the comprehensive performance score of the spaces is calculated. Based on the comprehensive spatial efficiency score, the space is divided into three levels: inefficient area (H<60), improvement area (60≤H<80), and demonstration area (H≥80), providing a quantitative basis for accurately identifying spatial problems.

[0008] Step S4: Identification of Inefficient Areas in Informal Residential Spaces Throughout the Day. Based on the evaluation system in Step S3, calculate the efficiency value of the informal residential space for each grid. According to the identified types of spatial inefficiency, construct a differentiated intervention scenario library for spatial inefficiency, including five categories: service support inefficiency, environmental safety inefficiency, spatial capacity inefficiency, economic vitality inefficiency, and ecological health inefficiency, as input conditions for simulation, accurately locating their spatial distribution and inefficiency types. This generates an efficiency map of informal residential spaces, visually displaying the spatial distribution and clustering characteristics of inefficient areas, and marking the dominant problem type for each inefficient area, providing targeted objectives for subsequent differentiated interventions.

[0009] Step S5: Construction of a Multi-Agent Model for Informal Residential Spaces. A multi-agent model of informal residential spaces is constructed, comprising resident agents, manager agents, and the physical environment. The resident agent's behavioral rules originate from the spatiotemporal behavioral patterns deconstructed in Step S1, including behavioral preferences, travel time, and travel mode. The manager agent represents the spatial planning and management department; its behavioral rules are based on the spatial intervention scenario library constructed in Step S4, enabling it to trigger corresponding spatial optimization measures in the simulation according to preset conditions. The physical environment is mapped from the spatial data of Steps S1 and S4, including roads, buildings, facilities, and inefficient areas. The time-bound functional suitability of the destination is evaluated; each spatial unit j possesses a time-bound functional spectrum based on its POI composition and the functional classification in Step S2. This identifies the main functions it can support at different times of the day, such as performing matching decisions. The decision of agent i for destination j is simplified to a binary selection function:

[0010] By adjusting the functional business format or management rules of the space, that is, changing To better match the all-day activity needs of the target audience, i.e., to respond , The purpose of the activity.

[0011] Step S6: Full-time optimization simulation of informal residential spaces based on multi-agent systems. The full-time dynamic simulation aims to eliminate inefficient areas and improve overall spatial efficiency. The differentiated spatial intervention scenario library constructed in Step S4 is used as input and imported into the multi-agent model constructed in Step S5 for dynamic simulation. This simulates the combined impact of different scenarios on residents' spatiotemporal behavior and the use of informal spaces at different times, dynamically outputting the full-time variation sequence of key indicators. The results are then imported back into the evaluation system of Step S3 for efficiency verification, dynamic efficiency assessment, and iterative optimization of the solution. Through multiple iterations of simulation, the optimal solution is output as an engineering report when the grid efficiency value for the entire area at all times is above 60 points.

[0012] Step S7: Visualization and Intelligent Control of Informal Residential Space Optimization Schemes. Optimization schemes are imported into a 3D interactive visualization platform for multi-scheme comparison and effect demonstration. Through user feedback and real-time monitoring data, model parameters and optimization strategies are dynamically adjusted to achieve continuous learning and iterative optimization of the schemes. A smart urban planning decision-making platform is constructed to achieve dynamic visualization and control of the schemes. The platform includes three modules: a scheme input and simulation module, used to import optimization schemes and drive multi-agent models for dynamic display; a multi-dimensional information visualization module, which overlays and displays current efficiency maps, optimization schemes, and simulated future efficiency comparisons in a 3D urban model; and a public participation and feedback module, which collects evaluation data from residents, experts, and management departments on the schemes through the platform, and establishes a dynamic optimization mechanism based on feedback information and monitoring data. When actual operating data deviates significantly from simulation predictions or new inefficiencies arise, the system will trigger an early warning, driving model parameter adjustments and scheme re-optimization, forming a "assessment-simulation-optimization-monitoring" intelligent planning closed loop.

[0013] Furthermore, the implementation process of step S1 is as follows: The aforementioned multi-source data collection and preprocessing of informal residential spaces in urban centers involves collecting multi-source heterogeneous data on informal spaces within the boundaries of a specific urban center. Spatial data collection includes urban boundary data, basic land use data, building outline data, road network data, and geometric and attribute data of the informal spaces themselves. For population dynamics data, mobile phone signaling data, WiFi probes, and video surveillance data are used. The mobile phone signaling data uses population dwell point information from any given seven days, including gender and age, with time accuracy better than one hour, and spatial positioning accuracy based on a 500m x 500m fine grid. WiFi probe sensors are deployed in key public spaces, building entrances and exits, and major circulation nodes within the study area. The spatial detection range is typically 50 meters, and the time resolution is to continuously scan and record device signals within the range at one-minute intervals. The addresses, signal strengths, and timestamps of nearby WiFi-enabled smart terminals are also continuously scanned and recorded to reconstruct individual movement trajectories, dwell time at key nodes, and movement speeds. Utilizing existing public safety surveillance cameras, traffic surveillance cameras, and potentially additional high-point panoramic cameras, continuous seven-day video streams were acquired of the study area. The video surveillance acquisition equipment had a resolution of 1080P (1920×1080 pixels) and a frame rate of 30 frames per second, enabling sub-second continuous behavior capture and the identification and collection of stationary, mobile, and clustered behavioral trajectories. A multi-target tracking algorithm was used to continuously track pedestrian movements across video frames, generating continuous spatiotemporal trajectory sequences. Spatial nodes where crowds gathered were automatically identified, and the crowd density and duration of these gatherings were calculated, generating a temporal variation curve of pedestrian traffic. The continuous data was then divided into four time periods: morning, noon, evening, and night. The acquired LBS and video stream data were preprocessed, with an anchoring time threshold of 10 minutes and a spatial threshold of 50 meters, removing abnormal and duplicate data. The filtered dataset was stored in a distributed database and a geospatial database to construct a spatiotemporal behavior database for informal residential spaces.

[0014] Furthermore, step S2 is implemented as follows: Step S2, the extraction of indicators for informal residential spaces in the urban center, mainly consists of three steps. The first step involves the division and identification of spatial functional units, using a 50m x 50m regular grid as the basic evaluation unit to systematically divide the study area. The second step extracts multidimensional characteristic variables from each grid unit. These physical spatial variables primarily include building function characteristics, business composition characteristics, population activity characteristics, and environmental background characteristics. Specifically, building function characteristics are the proportion of land area occupied by residential, commercial, and industrial buildings; business composition characteristics are the density and type distribution of various service facilities calculated based on POI data; population activity characteristics are obtained through LBS data, including the day-night population ratio, dwell time, and activity intensity; and environmental background characteristics are the green space ratio, water area, and proportion of impervious surfaces. These multidimensional characteristics are then systematically clustered, and based on the similarity of the functional attributes of the grid units, they are scientifically divided into three spatial categories: living-dominated, production-dominated, and ecological-dominated, laying the foundation for subsequent differentiated evaluation.

[0015] Furthermore, a comprehensive spatial evaluation index system was constructed based on the spatial function classification results. From the three core dimensions of living, production, and ecology, a multi-level evaluation index system was built, comprising 9 primary indicators and 16 secondary indicators. A four-time period division method was adopted: morning peak (7-9 am), daytime off-peak (9-5 pm), evening peak (5-7 pm), and nighttime (7 pm-7 am the next day). Weekends were analyzed separately. See the table below:

[0016] Furthermore, data standardization processing is performed. To eliminate the dimensional differences between indicators, an indicator standardization method is used to normalize the original data of all secondary indicators, linearly transforming them to the [0,1] interval. The specific calculation formula is as follows: ,in, For the indicator sample values, and These are the minimum and maximum values ​​of the indicator, respectively. This processing ensures the comparability of the indicator data, providing a standardized data foundation for subsequent spatial function evaluation.

[0017] Furthermore, step S3 is implemented as follows: The differentiated weighted evaluation system for the function orientation of informal residential spaces, constructed in step S3, dynamically presets weight sets for each of the three functional space types, assigning higher weights to indicators closely related to their core functions. Each evaluation unit uses the weight set corresponding to its space type to calculate its comprehensive performance score. In living-oriented spaces, service support and environmental safety are given core weights; in production-oriented spaces, informal employment density and functional mixing are given core weights; and in ecologically-oriented spaces, green space and open space coverage and blue-green network connectivity become the main evaluation criteria.

[0018] Furthermore, the system performs classification calculations and comprehensive evaluations, calculating the information entropy of each indicator. The specific formula is as follows If there are n indicators: j=1,2,...n, and i is the sample number, and if we are analyzing n grid regions, then i=1,2,...,n, where n is the total number of samples and m is the total number of grids. Let be the weight of the i-th sample on the j-th indicator. Each evaluation unit calculates its weight using the weight set of its spatial type. , Ultimately, a comprehensive evaluation result reflecting both universal applicability and functional characteristics is obtained. Based on this, a comprehensive performance evaluation model for informal residential spaces is constructed, and the comprehensive performance score is calculated. Based on the comprehensive spatial efficiency score, the space is divided into three levels: inefficient area (H<60), improvement area (60≤H<80), and demonstration area (H≥80), providing a quantitative basis for accurately identifying spatial problems.

[0019] Furthermore, step S4 is implemented as follows: The identification of inefficient areas in informal residential spaces in step S4 mainly consists of three steps. First, spatial efficiency calculation and classification: based on the differentiated evaluation model constructed in step S3, adapted to different spatial function types, the comprehensive efficiency value of each 50m × 50m grid unit of informal residential space is calculated. Based on the efficiency value, each grid unit is divided into three efficiency levels: inefficient areas (efficiency value < 60), improved areas (60 ≤ efficiency value < 75), and high-efficiency areas (efficiency value ≥ 75). Specifically, inefficient areas exhibit unbalanced spatial system operation, with significant functional shortcomings or serious problems, urgently requiring planning intervention; improved areas achieve basic spatial functions but have significant deficiencies in one or more aspects, possessing considerable potential for optimization and improvement; high-efficiency areas demonstrate coordinated spatial functions and efficient resource utilization, serving as exemplary models.

[0020] Furthermore, the causes and types of inefficient areas are analyzed, and a deeper diagnosis is conducted based on their spatial function types and scores of various indicators to analyze the root causes of inefficiency. These inefficiencies are categorized into five dominant types: service support inefficiency, environmental safety inefficiency, spatial capacity inefficiency, economic vitality inefficiency, and ecological health inefficiency. Service support inefficiency is mainly characterized by the coexistence of overloaded service facilities during peak hours and idle resources during off-peak hours, with a severe mismatch between public service hours and residents' activity patterns; this is commonly seen in residential-dominated spaces. Environmental safety inefficiency is characterized by the exacerbation of safety hazards during specific periods, such as insufficient nighttime lighting, constant obstruction of fire lanes, and a significant risk of flooding during the rainy season; this is common in both residential and production-dominated spaces, with varying characteristics during different periods. Spatial capacity inefficiency is mainly characterized by extreme fluctuations in space use during specific periods, with a stark contrast between overcrowding during peak hours and idle space at night; this is typically found in high-density residential-dominated spaces. Economic inefficiency is characterized by discontinuous economic activity across time periods, a lack of year-round business models, and short-lived commercial activity; it is a unique type of inefficiency specific to production-driven spaces. Ecological health inefficiency is mainly characterized by the temporal degradation of ecological services, with increased environmental pollution during specific periods and insufficient sustainability of ecological regulation functions; it is a major problem in ecology-driven spaces.

[0021] Furthermore, the three types of areas mentioned above are spatially located to generate a diagnostic map of the full-time efficiency and inefficiency types of informal residential spaces. This map visually displays the spatial distribution and clustering characteristics of inefficient areas and identifies the dominant problem type for each inefficient area. Based on the diagnostic map, a differentiated spatial intervention scenario library is constructed to clarify the correspondence between various inefficiency problems and specific planning intervention measures. For service-support inefficient areas, scenarios such as implanting micro-public service facilities and optimizing slow-traffic networks are pre-set; for environmental safety inefficient areas, scenarios such as clearing fire escape routes and promoting comprehensive management of dilapidated buildings are pre-set; and for economic vitality inefficient areas, scenarios such as guiding business upgrading and providing small-scale entrepreneurial spaces are pre-set. Based on this, a progressive analysis from spatial efficiency assessment to inefficiency cause analysis to optimization scenario pre-setting is achieved, closely linking macro-level spatial evaluation with micro-level planning intervention, providing accurate and structured input conditions for the simulation in step S6.

[0022] Furthermore, the implementation process of step S5 is as follows: The construction of the multi-agent model for informal residential spaces in step S5 involves the following steps: Based on the database, behavioral rules, and spatial diagnostic conclusions formed in the preceding steps, a multi-agent model for informal residential spaces is constructed to simulate the interaction between "people, environment, and management," providing a dynamic analysis platform for core inefficient space optimization. The model includes three core agents: resident agents, manager agents, and physical environment agents. Resident agents represent residents within informal residential spaces in the city center, and their initial attributes are initialized based on the statistical data and sample distribution generated from mobile signaling data from the preceding steps. Manager agents represent spatial planning and management departments, and their behavioral rules are based on the aforementioned spatial intervention scenario library, enabling them to trigger corresponding spatial optimization measures based on preset conditions during simulation. Physical environment agents serve as the activity carriers and objects of action for both residents and managers, and are mapped from the aforementioned spatiotemporal data and the generated spatial efficiency diagnostic map. Each grid cell contains its spatial function type and comprehensive efficiency value.

[0023] Furthermore, a multi-agent model with spatiotemporal activity pattern matching as the core decision-making logic is constructed. The activity intention for the current time period t is determined: each resident agent, based on the all-time activity pattern deconstructed in step S1, will automatically generate a dominant activity intention based on the current time during the simulation. The time-segment functional adaptability of the destination is evaluated. Each rule network j possesses a time-segment functional spectrum based on its POI composition and functional classification. It indicates the main functions it can support at different times of the day, such as a subway station entrance during the morning rush hour. For "efficient passage," the destination might become a "short stop" at midday. Executing the matching decision, agent i's decision regarding destination j simplifies to a binary choice function:

[0024] in: Whether the function of determining destination j in the current time period t supports the agent's activity intention. The core of this is to reflect the characteristics of all time periods, ensuring that the behavior of the intelligent agent changes dynamically over time. For distance threshold checking, i.e., the distance from i to j Is it less than the maximum tolerable distance allowed by the intent of this activity? The full-time dynamics of this model are entirely determined by... Activity Intent and Driven by two time-varying variables—space, time period, and function—the planning direction is clear: by adjusting the functional formats or management rules of the space, the needs of the target population for all-day activities can be better matched.

[0025] Furthermore, step S6 is implemented as follows: Step S6, the simulation of informal residential space optimization based on multi-agent systems, mainly consists of two steps: full-time dynamic simulation and deduction. With the overall goal of eliminating inefficient areas and improving overall spatial efficiency, the differentiated spatial intervention scenario library constructed in step S4 is used as input and imported into the multi-agent model constructed in step S5 for dynamic simulation. The simulation strictly follows full-time logic, simulating the comprehensive impact of different scenarios on residents' spatiotemporal behavior and informal space usage during typical periods such as morning peak, midday off-peak, evening peak, and nighttime. The model dynamically outputs the full-time change sequences of key indicators such as population thermal distribution, facility utilization rate, path traffic load, and activity type distribution, thereby proactively predicting potential spatial conflicts, traffic congestion, and new efficiency bottlenecks that planning interventions may trigger.

[0026] Furthermore, dynamic performance evaluation and iterative optimization involve importing the steady-state spatial state after each simulation round back into the evaluation system with differentiated weighting constructed in step S3. This recalculates the comprehensive performance value of each grid cell across all time periods throughout the day, achieving a quantitative translation and dynamic performance evaluation from "behavioral simulation" to "spatial performance." Through multiple iterative cycles of "scenario simulation - all-time evaluation," the system comprehensively compares the performance of each scheme in improving inefficient areas, maintaining the stability of efficient areas, and balancing the needs of different time periods. Ultimately, based on the core criterion of achieving no inefficient areas across the entire domain (performance value ≥ 60 points) and achieving Pareto optimality in the overall performance distribution, the recommended optimization scheme with the highest comprehensive benefits is selected.

[0027] Furthermore, step S7 is implemented as follows: The visualization and intelligent control of the informal residential space optimization scheme in step S7 mainly consists of two steps. The first step involves constructing a planning digital twin platform with interactive simulation and visualization modules. This seamlessly integrates and overlays the current status quo map, planning scheme, and future projections within a 3D city model. The current status quo map includes spatial data from step S1 and the current status efficiency map from step S4. The planning scheme includes the optimized spatial layout and facility configuration. By driving a multi-agent model, the system dynamically displays the simulation results of population activities and spatial efficiency throughout the entire time frame after the scheme's implementation, enabling the traceability of the planning scheme's effectiveness and future projections. The crowdsourcing negotiation and feedback module provides an open interface for the public, experts, and management departments. Through VR immersive experiences, scheme evaluation questionnaires, and interactive map annotation, it collects qualitative feedback and modification suggestions, incorporating subjective value judgments into the decision-making process.

[0028] Furthermore, a dynamic monitoring and intelligent evolution mechanism is established. The core intelligence of the platform lies in its dynamic closed loop of "evaluation-simulation-optimization-monitoring." After the implementation of the solution, the system continuously receives the real-time monitoring data stream described in step S1, automatically comparing the actual operating status with the simulation prediction results of step S6. When the system identifies a significant deviation between the actual operating data and the simulation prediction, or detects the generation of new inefficient areas through the evaluation system in step S3, it will automatically trigger a planning warning. This warning will drive the system to backtrack to step S4 or step S6, initiating a new round of bottleneck diagnosis, scenario generation, and simulation, thereby dynamically adjusting model parameters and optimization strategies. This enables the planning scheme to learn itself, iterate continuously, and optimize spirally, ultimately promoting the governance of informal residential spaces from the traditional "ultimate blueprint" model to adaptive, sustainable, and precise planning that adapts to the complex urban system.

[0029] Beneficial effects of this invention: (1) The accuracy of the evaluation of informal residential spaces in urban centers has increased from 60% to 90%. Due to limitations in data collection technology, traditional information acquisition mainly relies on manual surveys, questionnaire statistics, and simple data analysis methods, which have problems such as strong subjectivity, limited coverage, and insufficient accuracy, making it difficult to support the analysis needs of the current complex urban environment. Based on multi-source heterogeneous big data, this invention selects three dimensions—life dimension indicator system, production dimension indicator system, and ecological dimension indicator system—to analyze the population characteristic identification module and built environment characterization module of the model, comprehensively analyze the behavioral characteristics of the population in informal spaces in urban centers, and accurately identify the influence of static built environment factors and dynamic population behavior. Furthermore, for the division of spatial functional units and type identification, a 50m × 50m regular grid is used as the basic evaluation unit to systematically divide the study area. For each grid unit, multi-dimensional feature variables are extracted. The physical space variables mainly include building function characteristics, business format characteristics, population activity characteristics, and environmental background characteristics. Based on the spatial function classification results, a multi-level evaluation index system was constructed from three core dimensions: living, production, and ecology. This system includes 9 primary indicators and 16 secondary indicators, which improved the evaluation accuracy from 60% to 90%.

[0030] (2) The efficiency of optimizing the layout of informal residential spaces in urban centers is improved by 70%. Traditional optimization of informal spaces in urban centers mainly relies on the past experience of planners, combined with on-site surveys to derive optimization strategies, a process that is time-consuming and labor-intensive. This invention innovatively introduces a multi-agent simulation model, supplemented by multi-source big data collection technology, and uses a data-driven multi-agent simulation model to simulate informal spaces in urban centers. By changing 9 primary indicators and 16 secondary indicators in the three core dimensions of the living dimension indicator system, the production dimension indicator system, and the production dimension indicator system, the simulation evolution of informal residential spaces is promoted, thereby improving the efficiency of optimizing the layout of informal residential spaces in urban centers.

[0031] (3) Intelligent regulation and control of the layout optimization of informal residential spaces in urban centers. Traditional barrier-free space layout schemes cannot be automatically identified and regulated after generation. Intelligence relies on manual intervention, and the adjustment and optimization of the schemes require a large amount of repetitive manual labor, making it difficult to quickly respond to new needs or problems. Moreover, the decision-making process is not transparent, and the collaboration efficiency between relevant departments is low. This invention realizes the visualization and self-iteration of the optimization scheme for informal residential spaces by constructing an intelligent system that integrates a planning digital twin platform and a dynamic closed loop of "evaluation-simulation-optimization-monitoring". Its core lies in using digital twin technology to integrate multi-source data and conduct future scenario simulations, while opening it to the public and experts to collect feedback. After the scheme is implemented, it can automatically trigger early warning and optimization cycles through real-time data comparison, thereby promoting a fundamental transformation of space governance from a static "ultimate blueprint" model to a sustainable, intelligent, and adaptive precise planning paradigm. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the method implemented in this invention; Figure 2 This is a flowchart illustrating the multi-agent simulation process according to an embodiment of the present invention. Figure 3 This is a diagnostic chart of all-time performance and inefficiency types in an embodiment of the present invention; Figure 4 This is a diagram illustrating an optimized scheme according to an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0034] like Figure 1-4 As shown, a full-time simulation optimization method for informal residential spaces in the central area based on multi-agent systems is presented. This includes the following steps: I. Within the boundaries of Nanjing's central area, staff collected multi-source heterogeneous data on informal spaces. Spatial data included urban boundary data, basic land use data, building outline data, road network data, and geometric attribute data of the informal spaces themselves. For population dynamics data, mobile phone signaling data, WiFi probe data, and video surveillance data were used to identify and collect information on population dwelling, movement, and gathering behavior trajectories throughout any given seven days. The all-time data was divided into four time periods: morning, noon, evening, and night. The acquired population dynamics data underwent preprocessing, with a 10-minute anchoring time threshold and a 50-meter spatial threshold, removing abnormal and duplicate data. The resulting datasets were stored in a distributed database and a geospatial database, constructing a full-time spatiotemporal behavior database for informal residential spaces.

[0035] II. Based on the database from Step 1, staff members divided the study area into spatial functional units and identified their types. The study area was divided into basic grids of 50m x 50m. For each grid unit, multidimensional feature data was extracted. Based on its dominant function, each assessment unit was classified as a living-dominated, production-dominated, or ecology-dominated space, and multivariate cluster analysis was performed. A multi-level comprehensive evaluation index system for informal residential spaces was constructed from the three dimensions of living, production, and ecology, including 9 primary indicators and 16 secondary indicators. Finally, the index dataset was integrated, different indicator dimensions were standardized, and all data were linearly transformed to the 0-100% range using a standardized formula to achieve data normalization.

[0036] Third, based on the indicator dataset established in step two, perform classification calculations and comprehensive evaluations. Staff calculate the information entropy of each indicator. The specific formula is as follows If there are n indicators: j=1,2,...n, and i is the sample number, and if we are analyzing n grid regions, then i=1,2,...,n, where n is the total number of samples and m is the total number of grids. Let be the weight of the i-th sample on the j-th indicator; dynamically construct a differentiated weight set, and use the weight set corresponding to its spatial type for each evaluation unit, assigning higher weights to indicators closely related to the core functions, and calculate its comprehensive performance score to obtain the weight. Based on this, a comprehensive performance evaluation model for informal residential spaces is constructed, and the comprehensive performance score of the spaces is calculated. Based on the comprehensive spatial efficiency score, the space is divided into three levels: inefficient area (H<60), improvement area (60≤H<80), and demonstration area (H≥80), providing a quantitative basis for accurately identifying spatial problems.

[0037] Fourth, based on the evaluation system in step three, staff calculate the efficiency value of informal residential space for each grid. According to the identified types of spatial inefficiency, a differentiated intervention scenario library for spatial inefficiency is constructed, including five categories: service support inefficiency, environmental safety inefficiency, spatial capacity inefficiency, economic vitality inefficiency, and ecological health inefficiency. These serve as input conditions for simulation, accurately locating their spatial distribution and type of inefficiency. This generates an informal residential space efficiency map, visually displaying the spatial distribution and clustering characteristics of inefficient areas, and marking the dominant problem type for each inefficient area, providing targeted objectives for subsequent differentiated interventions.

[0038] V. Staff construct a multi-agent model of informal residential space, comprising resident agents, manager agents, and the physical environment. The resident agent's behavioral rules originate from the spatiotemporal behavioral patterns deconstructed in Step 1, including behavioral preferences, travel time, and travel mode. The manager agent represents the spatial planning and management department; its behavioral rules are based on the spatial intervention scenario library constructed in Step 4, enabling it to trigger corresponding spatial optimization measures in the simulation according to preset conditions. The physical environment is mapped from the spatial data of Steps 1 and 4, including roads, buildings, facilities, and inefficient areas. The time-segment functional suitability of the destination is assessed; each spatial unit j possesses a time-segment functional spectrum based on its POI composition and the functional classification in Step S2. This identifies the main functions it can support at different times of the day, such as performing matching decisions. The decision of agent i for destination j is simplified to a binary selection function:

[0039] By adjusting the functional business format or management rules of the space, that is, changing To better match the all-day activity needs of the target audience, i.e., to respond . VI. Staff members conduct full-time dynamic simulations to eliminate inefficient areas and improve overall spatial efficiency. Using the differentiated spatial intervention scenario library constructed in step four as input, they import it into the multi-agent model constructed in step five for dynamic simulation. This simulates the combined impact of different scenarios on residents' spatiotemporal behavior and informal space usage at different times, dynamically outputting the full-time variation sequence of key indicators. The results are then imported back into the evaluation system of step three for efficiency verification, dynamic efficiency assessment, and iterative optimization of the solution. Through multiple iterative simulations, the optimal solution is output as an engineering report when the grid efficiency value for the entire area at all times is above 60 points.

[0040] VII. Staff will import the optimized solutions into a 3D interactive visualization platform for multi-solution comparison and effect demonstration. Through user feedback and real-time monitoring data, model parameters and optimization strategies will be dynamically adjusted to achieve continuous learning and iterative optimization of the solutions. A smart decision-making platform for urban planning will be constructed to achieve dynamic visualization and control of the solutions. The platform includes three modules: a solution input and simulation module, used to import optimized solutions and drive the dynamic display of multi-agent models; a multi-dimensional information visualization module, which overlays and displays the current efficiency map, optimized solutions, and a comparison of simulated future efficiency in the 3D city model; and a public participation and feedback module, which collects evaluation data from residents, experts, and management departments on the solutions through the platform, and establishes a dynamic optimization mechanism based on feedback information and monitoring data. When actual operating data deviates significantly from simulation predictions or new inefficiencies arise, the system will trigger an early warning, driving model parameter adjustments and further optimization of the solutions, forming a closed-loop intelligent planning system of "evaluation-simulation-optimization-monitoring". Example

[0041] The technical solution of this invention will be described in detail below using the Huangjiawei shantytown in Nanjing, Jiangsu Province as an example.

[0042] (1) Within the Huangjiawei shantytown area of ​​Nanjing City, Jiangsu Province, staff collected urban spatial data and population dynamic data for any consecutive seven days.

[0043] (1.1) The anchoring time threshold is 10 minutes and the spatial threshold is 50 meters. Abnormal and duplicate data are removed. The filtered dataset is stored in a distributed database and a geographic database. (1.2) The static basic data includes urban boundary data, basic land use data, building outline data, road network data, and geometric and attribute data of informal spatial ontology. Specifically, the boundary data, building outline data, geometric and attribute data of informal spatial ontology, and urban basic land use data of Huangjiawei shantytown in Nanjing, Jiangsu Province, are all sourced from publicly available data from the Nanjing Municipal Government. Urban road data is collected through the OpenStreetMap platform. Regarding dynamic population data, this includes mobile phone signaling data, WiFi probe data, and video surveillance data. The mobile phone signaling data uses population dwell point information from January 6th to January 12th, 2025, including gender and age, with a time accuracy of 1 hour and spatial positioning accuracy based on a 500m*500m fine grid. WiFi probe sensors were deployed in key public spaces, building entrances and exits, and major traffic flow nodes within the Huangjiawei shantytown. The spatial detection range is typically 50 meters, and the sensors continuously scan and record device signals within the range at 1-minute intervals. They also continuously scan and record the addresses, signal strengths, and timestamps of nearby WiFi-enabled smart terminals, reconstructing individual movement trajectories, dwell time at key nodes, and movement speeds. Using existing urban public safety surveillance cameras, traffic surveillance cameras, and potentially additional high-point panoramic cameras, continuous seven-day surveillance video streams of the study area were acquired. The video surveillance acquisition equipment has a resolution of 1080P and a frame rate of 30 frames per second, enabling sub-second continuous behavior capture and identification and collection of information on stationary, flowing, and gathering behaviors. Through multi-target tracking algorithms, pedestrian movement trajectories were continuously tracked across video frames, generating continuous spatiotemporal trajectory sequences. Spatial nodes where crowds gather were automatically identified, and the crowd density and duration of the gathering areas were calculated, generating a temporal change curve for pedestrian traffic. The data was then divided into four time periods: morning, noon, evening, and night. The acquired LBS and video stream data were preprocessed, with an anchoring time threshold of 5 minutes and a spatial threshold of 30 meters, to remove abnormal and duplicate data. The resulting dataset was stored in a distributed database and a geospatial database to construct a spatiotemporal behavior database for informal residential spaces.

[0044] (2) Within the Huangjiawei shantytown area of ​​Nanjing City, Jiangsu Province, the staff extracted indicators for the Huangjiawei shantytown area in three main steps. The first step involved dividing and identifying spatial functional units, using a 10m × 10m regular grid as the basic evaluation unit to systematically divide the study area. The second step involved extracting multidimensional feature variables for each grid unit. The physical space variables mainly included building function characteristics, business composition characteristics, population activity characteristics, and environmental background characteristics. Specifically, building function characteristics were the proportion of land area occupied by residential, commercial, and industrial buildings; business composition characteristics were the density and type distribution of various service facilities calculated based on POI data; population activity characteristics were the day-night population ratio, length of stay, and activity intensity obtained through LBS data; and environmental background characteristics were the green space ratio, water area, and proportion of impermeable ground. The above multidimensional features were systematically clustered, and based on the similarity of the functional attributes of the grid units, they were divided into three spatial categories: living-dominated, production-dominated, and ecological-dominated, laying the foundation for subsequent differentiated evaluation.

[0045] (2.1) Staff members constructed a comprehensive spatial evaluation index system. Based on the spatial function classification results, a multi-level evaluation index system was constructed from three core dimensions: living, production, and ecology. The system includes 9 primary indicators and 16 secondary indicators, as shown in the table below: (2.2) The staff further standardized the data. To eliminate dimensional differences between indicators, an indicator standardization method was used to normalize the original data of all secondary indicators, linearly transforming them to the [0,1] interval. The specific calculation formula is as follows: ,in, For the indicator sample values, and These are the minimum and maximum values ​​of the indicator, respectively. This processing ensures the comparability of the indicator data, providing a standardized data foundation for subsequent spatial function evaluation.

[0046] (3) The staff constructed a differentiated weight evaluation system for the functional orientation of Huangjiawei shantytown, dynamically preset weight sets for the three functional space types, and assigned higher weights to indicators closely related to their core functions. Each evaluation unit used the weight set corresponding to its space type to calculate its comprehensive performance score. In the living-oriented space, service support and environmental safety were given core weights; in the production-oriented space, informal employment density and functional mixing were given core weights; and in the ecological-oriented space, green space and open space coverage and blue-green network connectivity became the main evaluation criteria.

[0047] (3.1) Staff members perform classification calculations and comprehensive evaluations of the above indicators, and calculate the information entropy of each indicator. The specific formula is as follows If there are n indicators: j=1,2,...n, and i is the sample number, and if we are analyzing n grid regions, then i=1,2,...,n, where n is the total number of samples and m is the total number of grids. Let be the weight of the i-th sample on the j-th indicator. Each evaluation unit calculates its weight using the weight set of its spatial type. , Ultimately, a comprehensive evaluation result reflecting both universal applicability and functional characteristics is obtained. Based on this, a comprehensive performance evaluation model for informal residential spaces is constructed, and the comprehensive performance score is calculated. Based on the comprehensive spatial efficiency score, the space is divided into three levels: inefficient area (H<60), improvement area (60≤H<80), and demonstration area (H≥80), providing a quantitative basis for accurately identifying spatial problems.

[0048] (4) The staff identified inefficient areas in the Huangjiawei shantytown by first calculating and classifying spatial efficiency. Based on the differentiated evaluation model constructed above and adapted to different spatial function types, the comprehensive efficiency value of the informal residential space in each 10m × 10m grid unit was calculated. According to the efficiency value, each grid unit was divided into the following three efficiency levels: inefficient area (efficiency value < 60), improved area (60 ≤ efficiency value < 75), and high-efficiency area (efficiency value ≥ 75). Finally, it was found that there were a large number of inefficient areas in the Huangjiawei shantytown, accounting for 80% of the total area, improved areas accounting for 17% of the total area, and high-efficiency areas accounting for 3% of the total area. Specifically, inefficient regional spatial systems are unbalanced in operation, with significant functional shortcomings or serious problems. The buildings are mostly shantytowns and old buildings, posing safety hazards and functional deficiencies, and urgently need to be demolished and renovated. The basic functions of regional spaces have been improved, but their environmental carrying capacity and economic vitality are clearly insufficient, and they have great potential for optimization and improvement. Efficient regional spaces have coordinated functions and efficient resource utilization, and can serve as demonstration models.

[0049] (4.1) Staff members analyzed the causes and classified the inefficient areas, and further conducted in-depth diagnosis based on their spatial function types and scores of various indicators to analyze the root causes of inefficiency. These were categorized into five dominant types: service support inefficiency, environmental safety inefficiency, spatial capacity inefficiency, economic vitality inefficiency, and ecological health inefficiency. Service support inefficiency is mainly characterized by overloaded service facilities during peak hours and idle resources during off-peak hours, with a serious mismatch between public service hours and residents' activity patterns. This is commonly seen in residential-dominated spaces. Environmental safety inefficiency is characterized by a periodic increase in safety hazards, including insufficient nighttime lighting, constant occupation of fire lanes, and a prominent risk of flooding during the rainy season. This is common in both residential and production-dominated spaces, with varying risk periods. Spatial capacity inefficiency is mainly characterized by extreme fluctuations in space use during peak hours, with a stark contrast between overcrowding during peak hours and idle space at night. This is typically found in high-density residential-dominated spaces. Economic inefficiency is characterized by discontinuous economic activity across time periods, a lack of year-round business models, and short-lived commercial activity; it is a unique type of inefficiency specific to production-driven spaces. Ecological health inefficiency is mainly characterized by the temporal degradation of ecological services, with increased environmental pollution during specific periods and insufficient sustainability of ecological regulation functions; it is a major problem in ecology-driven spaces.

[0050] (4.2) Staff spatially located the five dominant types of areas mentioned above, generating a diagnostic map of the Huangjiawei shantytown's all-time efficiency and inefficiency types. This map visually displays the spatial distribution and clustering characteristics of inefficient areas and marks the dominant problem type for each inefficient area. Based on the diagnostic map, a differentiated spatial intervention scenario library was constructed to clarify the correspondence between various inefficiency problems and specific planning intervention measures. For service-support inefficient areas, scenarios such as implanting micro-public service facilities and optimizing the slow-traffic network were preset; for environmental safety inefficient areas, scenarios such as opening up fire escape routes and promoting comprehensive management of dilapidated buildings were preset; and for economic vitality inefficient areas, scenarios such as guiding business upgrading and providing small-scale entrepreneurial space were preset. Based on this, a progressive analysis from spatial efficiency assessment to inefficiency cause analysis to optimization scenario preset was achieved, closely linking macro-level spatial evaluation with micro-level planning intervention, providing accurate and structured input conditions for the all-time optimization simulation of the Huangjiawei shantytown.

[0051] (5) Staff members constructed a multi-agent model for the Huangjiawei shantytown. The specific steps were as follows: based on the database, behavioral rules, and spatial diagnostic conclusions formed in the aforementioned steps, a multi-agent model of informal residential space capable of simulating the interaction between "people-environment-management" was constructed to provide a dynamic analysis platform for the optimization of core inefficient spaces. The model includes three types of core agents: resident agents, manager agents, and physical environment agents. The resident agents are the residents in the Huangjiawei shantytown, and their initial attributes are initialized based on the sample distribution generated from the statistical data and mobile phone signaling data in the aforementioned steps. The manager agents represent the spatial planning and management departments, and their behavioral rules are based on the aforementioned spatial intervention scenario library, enabling them to trigger corresponding spatial optimization measures in the simulation according to preset conditions. The physical environment agents, as the activity carriers and objects of action for residents and managers, are mapped by the aforementioned spatiotemporal data and the generated spatial efficiency diagnostic map. They include the functions and boundaries of buildings, roads, and plots in the Huangjiawei shantytown. Each grid cell in the spatial efficiency diagnostic map contains its spatial function type and comprehensive efficiency value.

[0052] (5.1) The staff constructs a multi-agent model with spatiotemporal activity pattern matching as the core decision-making logic. Determine the activity intention of the current time period t: Each resident agent will automatically generate a dominant activity intention based on the current time according to the full-time activity pattern deconstructed in step (1) during the simulation process. The time-slot functional suitability of the destination is assessed. Each basic assessment unit j possesses a time-slot functional spectrum based on its POI composition and the functional classification in step S2. This identifies the main functions it can support at different times of the day. Executing a matching decision, agent i's decision regarding destination j simplifies to a binary selection function:

[0053] in: Whether the function of determining destination j in the current time period t supports the agent's activity intention. This ensures that the behavior of the intelligent agent changes dynamically over time. For distance threshold checking, i.e., the distance from i to j Is it less than the maximum tolerable distance allowed by the intent of this activity? The full-time dynamics of this model are entirely determined by... Activity Intent and Driven by two time-varying variables—space, time period, and function—the planning direction is clear: by adjusting the functional formats or management rules of the space, the needs of the target population for all-day activities can be better matched.

[0054] (6) Staff members constructed a simulation model for optimizing the Huangjiawei shantytown based on multi-agent systems. This model mainly consisted of two steps. The first step was a full-time dynamic simulation, with the overall goal of eliminating inefficient areas and improving overall spatial efficiency. The aforementioned differentiated spatial intervention scenario library was used as input conditions and imported into the multi-agent model for dynamic simulation. The simulation strictly followed the full-time logic, simulating the comprehensive impact of different scenarios on residents' spatiotemporal behavior and informal space usage during typical periods such as morning peak, midday off-peak, evening peak, and nighttime. The model dynamically output the full-time change sequence of key indicators such as population thermal distribution, facility utilization rate, path traffic load, and activity type distribution, thereby proactively predicting potential spatial conflicts, traffic congestion, and new efficiency bottlenecks that may be caused by planning interventions.

[0055] (6.1) Staff conduct dynamic performance evaluation and iterative optimization of the scheme. After each round of simulation, the steady-state spatial state is re-imported into the evaluation system with differentiated weights. The comprehensive performance value of each basic evaluation unit in all time periods throughout the day is recalculated, realizing the quantitative translation from behavioral simulation to spatial performance and dynamic performance evaluation. Through multiple iterations of "scenario simulation-all-time evaluation", the system will comprehensively compare the performance of each scheme in improving inefficient areas, maintaining the stability of efficient areas, and balancing the needs of different time periods. Finally, based on the core criterion of achieving no inefficient areas in the entire domain and achieving Pareto optimality in the overall performance distribution, the recommended optimization scheme with the highest comprehensive benefits is selected.

[0056] (7) Staff members visualized and intelligently controlled the optimization plan for the Huangjiawei shantytown, constructed a planning digital twin platform, and seamlessly integrated and overlaid the current status map, planning scheme, and future projections in the three-dimensional city model through interactive simulation and visualization modules. The current status map includes spatial data and current efficiency map, and the planning scheme includes the optimized spatial layout and facility configuration. By driving the multi-agent model, the simulation results of population activities and spatial efficiency at all times after the implementation of the scheme are dynamically displayed, realizing the efficiency traceability and future projection of the planning scheme. The crowdsourcing negotiation and feedback module is an open interface for the public, experts and management departments. Through VR immersive experience, scheme evaluation questionnaires and interactive map annotation, qualitative feedback and modification suggestions are collected, and subjective value judgments are incorporated into the decision-making process.

[0057] (7.2) Establish a dynamic monitoring and intelligent evolution mechanism. The core intelligence of the platform lies in its dynamic closed loop of "evaluation-simulation-optimization-monitoring". After the implementation of the scheme, the system continuously connects to the real-time monitoring data stream and automatically compares the actual operating status with the simulation prediction results. When the system identifies a significant deviation between the actual operating data and the simulation prediction, or detects the generation of new inefficient areas through the evaluation system, it will automatically trigger a planning warning. This warning will drive the system to retrospectively start a new round of bottleneck diagnosis, scenario generation and simulation, thereby dynamically adjusting the model parameters and optimization strategies, realizing the self-learning, continuous iteration and spiral optimization of the planning scheme, and ultimately promoting the governance of informal residential spaces from the traditional "ultimate blueprint" model to adaptive, sustainable and precise planning that adapts to the complex urban system.

Claims

1. A multi-agent-based full-time simulation optimization method for informal residential spaces in the central area, comprising the following steps: Step S1: Full-time multi-source data collection and preprocessing of informal residential spaces Within the boundary of a city center, multi-source heterogeneous data of informal spaces are collected. In terms of spatial data, urban boundary data, basic land use data, building outline data, road network data, and geometric attribute data of the informal space itself are collected. In terms of population dynamic data, information on the trajectory of population residence, movement, and gathering behavior at all times within any seven days is identified and collected through mobile phone signaling data, WiFi probes, and video surveillance data. The full-time data is divided into four time periods: morning peak, off-peak, evening peak, and night. The acquired dynamic data of the population is preprocessed with a time threshold of 10 minutes and a spatial threshold of 50 meters to remove abnormal and duplicate data. The dataset after the above filtering is stored in a distributed database and a geospatial database to construct a full-time spatiotemporal behavior database of informal residential spaces. Step S2: Functional Identification and Indicator Extraction of Informal Residential Spaces in Urban Centers Based on the database from step S1, the study area is divided into spatial functional units and identified by type. The study area is divided into basic grids of 50m × 50m. For each grid unit, multidimensional feature data is extracted. Based on its dominant function, each evaluation unit is divided into living-dominated, production-dominated, or ecology-dominated spaces, and multivariate cluster analysis is performed. From the three dimensions of living, production, and ecology, a multi-level comprehensive evaluation index system for informal residential spaces is constructed, including 9 primary indicators and 16 secondary indicators. Finally, the index dataset is integrated to unify the different indicator dimensions. All data are linearly transformed to the range of 0 to 100% using a standardized formula to achieve data normalization. Step S3: Construction of a Differentiated Weighting Evaluation System for Informal Residential Spaces Based on Functional Orientation Based on the indicator dataset established in step S2, classification calculations and comprehensive evaluations are performed; the information entropy of each indicator is calculated. The specific formula is as follows If there are n indicators: j=1,2,...n, and i is the sample number, and if we are analyzing n grid regions, then i=1,2,...,n, where n is the total number of samples and m is the total number of grids. Let be the weight of the i-th sample on the j-th indicator; dynamically construct a differentiated weight set, and use the weight set corresponding to its spatial type for each evaluation unit, assigning higher weights to indicators closely related to the core functions, and calculate its comprehensive performance score to obtain the weight. Based on this, a comprehensive performance evaluation model for informal residential spaces is constructed, and the comprehensive performance score of the spaces is calculated. Let H be the score of the j-th indicator; based on the comprehensive spatial efficiency score, the space is divided into three levels: inefficient area (H<60), improvement area (60≤H<80), and demonstration area (H≥80), providing a quantitative basis for accurately identifying spatial problems; Step S4: Identification of Inefficient Areas in Informal Living Spaces Throughout the Day Based on the evaluation system in step S3, the efficiency value of informal residential space for each grid is calculated. According to the identified types of spatial inefficiency, a differentiated intervention scenario library for spatial inefficiency is constructed, including five categories: service support inefficiency, environmental safety inefficiency, spatial capacity inefficiency, economic vitality inefficiency, and ecological health inefficiency, which serve as input conditions for simulation. This accurately locates their spatial distribution and type of spatial inefficiency. Furthermore, an informal residential space efficiency map is generated, visually displaying the spatial distribution and clustering characteristics of inefficient areas, and marking the dominant problem type for each inefficient area, providing targeted objectives for subsequent differentiated interventions. Step S5: Construction of a multi-agent model for informal living spaces throughout all times Construct a multi-agent model of informal residential space that includes resident intelligent agents, manager intelligent agents, and physical space environment; the behavioral rules of resident intelligent agents are derived from the spatiotemporal behavioral patterns deconstructed in step S1, including behavioral preferences, travel time, and travel mode; the manager intelligent agent represents the spatial planning and management department, and its behavioral rules are based on the spatial intervention scenario library constructed in step S4, which can trigger corresponding spatial optimization measures in the simulation according to preset conditions. The physical spatial environment is mapped from the spatial data in steps S1 and S4, including roads, buildings, facilities, and inefficient areas; the time-bound functional suitability of the destination is assessed, and each spatial unit j has a time-bound functional spectrum based on its POI composition and the functional classification in step S2. This identifies the main functions it can support at different times of the day, such as performing matching decisions. The decision of agent i for destination j is simplified to a binary selection function: ， By adjusting the functional business format or management rules of the space, that is, changing To better match the all-day activity needs of the target audience, i.e., to respond , For the purpose of the activity; Step S6: Full-time optimization simulation of informal living spaces based on multi-agent systems The full-time dynamic simulation and deduction aims to eliminate inefficient areas and improve the overall spatial efficiency. The differentiated spatial intervention scenario library constructed in step S4 is used as input conditions and imported into the multi-agent model constructed in step S5 for dynamic simulation. It simulates the comprehensive impact of different scenarios on residents' spatiotemporal behavior and informal space use at different times, and dynamically outputs the full-time change sequence of key indicators. The results are then imported into the evaluation system of step S3 for efficiency verification, dynamic efficiency evaluation and scheme iteration optimization. Through multiple iterations of simulation, the scheme stops when the grid efficiency value of the entire area at all times is above 60 points. The optimal scheme is then output as an engineering report. Step S7: Visualization and Intelligent Control of Informal Living Space Optimization Solutions The optimized solutions are imported into a 3D interactive visualization platform for comparison and effect display of multiple solutions; through user feedback and real-time monitoring data, model parameters and optimization strategies are dynamically adjusted to achieve continuous learning and iterative optimization of the solutions; a smart decision-making platform for urban planning is built to realize dynamic visualization and control of the solutions; the platform includes three modules: a solution input and deduction module, which is used to import optimized solutions and drive multi-agent models to perform dynamic display. The multi-dimensional information visualization module overlays and displays current performance maps, optimization plans, and comparisons of future performance after simulation within a 3D city model. The public participation and feedback module collects evaluation data on the plan from residents, experts and management departments through the platform. Based on feedback information and monitoring data, a dynamic optimization mechanism is established. When there is a significant deviation between the actual operation data and the simulation prediction or when new inefficiencies occur, the system will trigger an early warning, drive the adjustment of model parameters and the re-optimization of the plan, and form a planning intelligent closed loop of "evaluation-simulation-optimization-monitoring".

2. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, In step S1, multi-source data collection and preprocessing of informal residential spaces in the urban center area throughout all time periods involves collecting heterogeneous data from multiple sources within the boundary of a specific urban center area. Spatial data collection includes urban boundary data, basic land use data, building outline data, road network data, and geometric and attribute data of the informal space itself. For population dynamics data, mobile phone signaling data, WiFi probes, and video surveillance data are used. Mobile phone signaling data uses population dwell point information from any seven-day period, including gender and age, with time accuracy better than one hour, and spatial positioning accuracy based on a 500m*500m fine grid. WiFi probe sensors are deployed in key public spaces, building entrances and exits, and major flow nodes within the study area. The spatial detection range is typically 50 meters, and the time resolution is to continuously scan and record device signals within the range at one-minute intervals. The addresses, signal strengths, and timestamps of nearby WiFi-enabled smart terminals are continuously scanned and recorded to reconstruct individual movement trajectories, dwell time at key nodes, and movement speeds. This data is then used to... Existing public safety surveillance cameras, traffic surveillance cameras, and potentially added high-point panoramic cameras in the city were used to acquire continuous seven-day full-time surveillance video streams of the study area. The video surveillance acquisition equipment had a resolution of 1080P and a frame rate of 30 frames per second, enabling sub-second continuous behavior capture and identification and collection of stationary, mobile, and gathering behavior trajectory information. Through a multi-target tracking algorithm, the movement trajectory of pedestrians was continuously tracked across video frames, generating a continuous spatiotemporal trajectory sequence, automatically identifying spatial nodes where crowds gathered, and calculating the crowd density and duration of the gathering area to generate a temporal change curve of pedestrian flow. The full-time data was then divided into four time periods: morning, noon, evening, and night. The acquired LBS and video stream data were preprocessed, with an anchoring time threshold of 10 minutes and a spatial threshold of 50 meters, and abnormal and duplicate data were removed. The dataset after the above filtering was stored in a distributed database and a geospatial database to construct a spatiotemporal behavior database of informal residential spaces.

3. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, In step S2, the extraction of indicators for informal residential spaces in the urban center first involves dividing the study area into functional units and identifying types. A regular grid of 50 meters × 50 meters is used as the basic evaluation unit. The second step involves extracting multidimensional feature variables for each grid unit. The physical space variables mainly include building function characteristics, business composition characteristics, population activity characteristics, and environmental background characteristics. The building function characteristics are the proportion of land area occupied by residential, commercial and industrial buildings; the business format characteristics are the density and type distribution of various service facilities calculated based on POI data; the population activity characteristics are the day-night population ratio, length of stay and activity intensity obtained through LBS data; the environmental background characteristics are the green space ratio, water area and impermeable ground ratio; the above multidimensional characteristics are systematically clustered, and based on the similarity of the functional attributes of grid units, they are scientifically divided into three major spatial categories: living-oriented, production-oriented and ecology-oriented, laying the foundation for subsequent differentiated evaluation; The spatial comprehensive evaluation index system is constructed based on the spatial function classification results. From the three core dimensions of living, production, and ecology, a multi-level evaluation index system is built, comprising 9 primary indicators and 16 secondary indicators. A four-time period division method is adopted: morning peak (7-9 am), daytime off-peak (9-5 pm), evening peak (5-7 pm), and nighttime (7-7 am the next day). Weekends are analyzed separately. See the table below: Data standardization processing: To eliminate the dimensional differences between indicators, an indicator standardization method is used to normalize the original data of all secondary indicators, linearly transforming them to the [0,1] interval; the specific calculation formula is as follows: ,in, For indicator sample values and These are the minimum and maximum values ​​of the indicator, respectively. After this processing, the data of each indicator are comparable, providing a standardized data foundation for subsequent spatial function evaluation.

4. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, In step S3, the construction of a differentiated weighted evaluation system for the function orientation of informal residential spaces involves dynamically pre-setting weight sets for the three functional space types, with higher weights assigned to indicators closely related to their core functions. Each evaluation unit uses the weight set corresponding to its space type to calculate its comprehensive performance score. In living-oriented spaces, service support and environmental safety are given core weights. In production-oriented spaces, informal employment density and functional mixing are given core weights. In ecological-oriented spaces, green space and open space coverage and blue-green network connectivity become the main evaluation criteria. Classification calculation and comprehensive evaluation, calculating the information entropy of each indicator. The specific formula is as follows If there are n indicators: j=1,2,...n, and i is the sample number, and if we are analyzing n grid regions, then i=1,2,...,n, where n is the total number of samples and m is the total number of grids. Let be the weight of the i-th sample on the j-th indicator; each evaluation unit uses the weight set of its spatial type to calculate the weight. , , Let the information entropy of the j-th indicator be used to obtain a comprehensive evaluation result that reflects the universality and functional characteristics. Based on this, a comprehensive performance evaluation model for informal residential spaces is constructed, and the comprehensive performance score of the space is calculated. The score for the j-th indicator is the specific value of each evaluation unit on the j-th indicator. Based on the comprehensive spatial efficiency score, the space is divided into three levels: inefficient area (H<60), improved area (60≤H<80), and demonstration area (H≥80), providing a quantitative basis for accurately identifying spatial problems.

5. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, The identification of inefficient areas in informal residential spaces in step S4 mainly consists of three steps. The first step is the calculation and classification of spatial efficiency. Based on the differentiated evaluation model adapted to different spatial function types constructed in step S3, the comprehensive efficiency value of informal residential spaces in each 50m×50m grid unit is calculated. According to the efficiency value, each grid unit is divided into the following three efficiency levels, including inefficient areas with efficiency values ​​below 60. Areas requiring improvement have an efficiency value between 60 and 75 (below 75), while high-efficiency areas have an efficiency value greater than 75. Inefficient areas exhibit an imbalance in the operation of the spatial system, with significant functional shortcomings or serious problems, and urgently require planning intervention. While the basic functions of the regional space have been improved, there are still significant shortcomings in one or more aspects, indicating considerable potential for optimization and improvement. The efficient coordination of regional spatial functions and the high efficiency of resource utilization serve as a model; The causes and types of inefficient areas were analyzed and classified. Further in-depth diagnosis was conducted based on their spatial function types and scores of various indicators to analyze the root causes of inefficiency. These inefficiencies were categorized into five dominant types: service support inefficiency, environmental safety inefficiency, spatial capacity inefficiency, economic vitality inefficiency, and ecological health inefficiency. Service support inefficiency is characterized by overloaded service facilities during peak hours and idle resources during off-peak hours, with a severe mismatch between public service hours and residents' activity patterns; this is commonly seen in residential-dominated spaces. Environmental safety inefficiency is characterized by a temporal increase in safety hazards, such as insufficient nighttime lighting, constant obstruction of fire lanes, and a prominent risk of flooding during the rainy season; this is common in both residential and production-dominated spaces, with different characteristics during different periods of risk. Inefficient spatial capacity is characterized by extreme fluctuations in space usage during peak hours, with a stark contrast between overcrowding during peak hours and idle space at night. This type is typically found in high-density living spaces. Economic inefficiency is characterized by time-series discontinuities in economic activity, a lack of year-round business formats, and short-lived commercial vitality, making it a unique type of inefficiency in production-driven spaces. Ecological health inefficiency is characterized by time-series degradation of ecological environment services, with environmental pollution worsening during specific periods and insufficient sustainability of ecological regulation functions, which is the main problem in ecologically driven spaces. The three types of areas mentioned above are spatially located to generate a diagnostic map of the full-time efficiency and inefficiency types of informal residential spaces. This map visually displays the spatial distribution and clustering characteristics of inefficient areas and marks the dominant problem type for each inefficient area. Based on the diagnostic map, a differentiated spatial intervention scenario library is constructed to clarify the correspondence between various inefficiency problems and specific planning intervention measures. For service-support inefficient areas, scenarios such as implanting micro-public service facilities and optimizing slow-moving traffic networks are preset. For environmental safety inefficient areas, scenarios such as opening up fire escape routes and promoting comprehensive management of dilapidated buildings are preset. For economic vitality inefficient areas, scenarios such as guiding business upgrading and providing small-scale entrepreneurial space are preset. Based on this, a progressive analysis from spatial efficiency assessment to inefficiency cause analysis to optimization scenario preset is achieved, closely linking macro-level spatial evaluation with micro-level planning intervention, providing accurate and structured input conditions for the simulation in step S6.

6. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, The construction of the multi-agent model for informal residential spaces in step S5 involves the following steps: Based on the database, behavioral rules, and spatial diagnostic conclusions formed in the preceding steps, a multi-agent model for informal residential spaces that can simulate the interaction between "people, environment, and management" is constructed to provide a dynamic analysis platform for the core optimization of inefficient spaces. The model includes three types of core agents: resident agents, manager agents, and physical environment agents. The resident agent represents residents in informal residential spaces in the city center, and its initial attributes are initialized based on the statistical data and sample distribution generated from mobile phone signaling data in the preceding steps. The manager agent represents the spatial planning and management department, and its behavioral rules are based on the aforementioned spatial intervention scenario library, enabling it to trigger corresponding spatial optimization measures in the simulation according to preset conditions. The physical environment agent serves as the activity carrier and object of action for both residents and managers, and is mapped from the aforementioned spatiotemporal data and the generated spatial efficiency diagnostic map. Each grid cell contains its spatial function type and comprehensive efficiency value. A multi-agent model with "spatiotemporal activity pattern matching" as the core decision-making logic is constructed; the activity intention of the current time period t is determined: each resident agent will automatically generate a dominant activity intention based on the current time according to the full-time activity pattern deconstructed in step S1 during the simulation. Assess the time-segment functional suitability of the destination. Each spatial unit j possesses a time-segment functional spectrum based on its POI composition and the functional classification in step S2. This identifies the main functions it can support at different times of the day; to perform matching decisions, the decision of agent i for destination j is simplified to a binary selection function: , in: Whether the function of determining destination j in the current time period t supports the agent's activity intention. This is the core feature that embodies the characteristics of all-time, ensuring that the behavior of the agent changes dynamically over time. For distance threshold checking, i.e., the distance from i to j Is it less than the maximum tolerable distance allowed by the intent of this activity? The full-time dynamics of this model are entirely determined by... Activity Intent and Driven by two time-varying variables—space, time period, and function—the planning direction is clear: by adjusting the functional formats or management rules of the space, the needs of the target population for all-day activities can be better matched.

7. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, The simulation of informal residential space optimization based on multi-agent technology in step S6 is divided into two steps: a full-time dynamic simulation and deduction, with the overall goal of eliminating inefficient areas and improving the overall spatial efficiency. The differentiated spatial intervention scenario library constructed in step S4 is used as input conditions and imported into the multi-agent model constructed in step S5 for dynamic simulation. The simulation strictly follows the logic of the entire time period, simulating the comprehensive impact of different scenarios on residents' spatiotemporal behavior and the use of informal spaces during typical periods of morning peak, midday off-peak, evening peak, and night. The model dynamically outputs the full-time change sequence of key indicators such as population thermal distribution, facility utilization rate, path traffic load, and activity type distribution, thereby proactively predicting spatial conflicts, traffic congestion, and new efficiency bottlenecks that may be caused by planning intervention. Dynamic performance evaluation and iterative optimization involve importing the steady-state spatial state after each simulation into the evaluation system with differentiated weights constructed in step S3. This recalculates the comprehensive performance value of each grid cell throughout the day, achieving a quantitative translation from "behavioral simulation" to "spatial performance" and dynamic performance evaluation. Through multiple iterative cycles of "scenario simulation - all-time evaluation," the system comprehensively compares the performance of each scheme in improving inefficient areas, maintaining the stability of efficient areas, and balancing the needs of different time periods. Finally, based on the core criterion of achieving no inefficient areas across the entire domain and Pareto optimality in the overall performance distribution, the recommended optimization scheme with the highest comprehensive benefits is selected.

8. The method for full-time simulation optimization of irregular residential spaces in the central area based on multi-agent systems according to claim 1, characterized in that, The visualization and intelligent control of the informal residential space optimization scheme in step S7 is divided into two steps: constructing a planning digital twin platform and an interactive simulation and visualization module, which seamlessly integrates and overlays the current status map, planning scheme, and future projections in a three-dimensional city model; the current status map includes spatial data based on step S1 and the current efficiency map in step S4; the planning scheme includes the optimized spatial layout and facility configuration; by driving a multi-agent model, the simulation results of crowd activities and spatial efficiency at all times after the implementation of the scheme are dynamically displayed, realizing the efficiency traceability and future projection of the planning scheme; the crowdsourcing negotiation and feedback module is an open interface for the public, experts, and management departments, which collects qualitative feedback and modification suggestions through VR immersive experience, scheme evaluation questionnaires, and interactive map annotation, incorporating subjective value judgments into the decision-making process; The platform establishes a dynamic monitoring and intelligent evolution mechanism. Its core intelligence lies in the dynamic closed loop of "evaluation-simulation-optimization-monitoring". After the implementation of the plan, the system continuously connects to the real-time monitoring data stream described in step S1 and automatically compares the actual operating status with the simulation prediction results in step S6. When the system identifies a significant deviation between the actual operating data and the simulation prediction, or detects the generation of new inefficient areas through the evaluation system in step S3, it will automatically trigger a planning warning. This warning will drive the system to backtrack to step S4 or step S6 and start a new round of bottleneck diagnosis, scenario generation and simulation, thereby dynamically adjusting model parameters and optimization strategies to achieve self-learning, continuous iteration and spiral optimization of the planning scheme. Ultimately, this will promote the governance of informal residential spaces from the traditional "ultimate blueprint" model to adaptive, sustainable and precise planning that adapts to the complex urban system.

Images