Community commercial facility space planning layout method based on urban multi-agent simulation

Abstract

The invention discloses a community commercial facility space planning and layout method based on city multi-agent simulation. Comprising the following steps: S1, multi-source data fusion and high-precision three-dimensional modeling; S2, multi-agent cooperative system construction and behavior rule implantation; S3, spatio-temporal behavior flow simulation engine operation; S4, community commercial supply and demand matching blind area diagnosis; through evaluation of a simulation verification system, the commercial service coverage efficiency is optimized by 72%, the matching degree of facilities and target crowds is increased by 64%, and dynamic decision support can be provided for community commercial planning.

Description

Technical Field

[0001] This invention belongs to the field of smart city planning technology, specifically relating to a method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation. Background Technology

[0002] With the accelerated advancement of my country's new urbanization process, the scientific layout of community commercial facilities directly impacts the quality of public services and urban spatial efficiency. Current mainstream planning methods still rely on static configuration based on census data and economic models, primarily using Geographic Information Systems (GIS) for static spatial overlay and buffer zone analysis, or employing methods like the Analytic Hierarchy Process (AHP) to construct multi-factor site selection models. These methods struggle to adapt to the complex needs of the dynamic evolution of high-density urban communities. Significant shortcomings of existing technologies are particularly evident in the renovation of old residential areas and the construction of age-friendly communities.

[0003] 1. Traditional two-dimensional analysis models cannot quantify the attenuation effect of three-dimensional terrain on walkability, resulting in distorted service radius calculations;

[0004] 2. The static indicator system lacks the ability to dynamically respond to residents' spatial and temporal behavior patterns and the competitive relationship among multiple business formats, resulting in a mismatch between commercial supply and demand;

[0005] 3. In the face of the accelerating trend of an aging society, the existing system has failed to establish a mandatory protection mechanism for vulnerable groups.

[0006] Existing technologies are insufficient to achieve a closed-loop decision-making process for the layout of community commercial facilities, from micro-level behavioral simulation to macro-level scheme optimization. There is an urgent need for a systematic solution that can integrate multi-source big data, establish precise behavioral rules, support real-time simulation and deduction, and output flexible optimization schemes. Summary of the Invention

[0007] To address the shortcomings mentioned in the background, this invention aims to provide a spatial planning and layout method for community commercial facilities based on urban multi-agent simulation. This method constructs a three-dimensional voxel model by fusing multi-source urban data, identifies service blind spots using supply-demand gap calculation and diagnostic techniques, generates facility layout schemes using a flexible adaptation optimization algorithm, and finally verifies the optimization results through simulation. This method is applicable to the scientific layout and dynamic optimization of urban community commercial facilities, and particularly provides a solution to the problem of optimizing the allocation of limited space resources in high-density communities.

[0008] The technical solution adopted in this invention is as follows:

[0009] A method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation is disclosed. The system employed by this method includes a user interface device, a data transmission network, and a cloud computing platform. The user interface device comprises a data acquisition unit, a 3D display unit, and a human-computer interaction unit. The data acquisition unit is used to acquire and input 3D point cloud data, multi-period land use attribute data, community population census and real-time updated data, social media sentiment analysis data, road network topology vector data, and commercial facility point of interest (POI) data. The 3D display unit is used for visual rendering of the planning scheme, and the human-computer interaction unit is used for parameter adjustment. The data transmission network uses an encrypted transmission channel to connect the user interface device and the cloud computing platform. The cloud computing platform includes a data fusion engine, a multi-agent simulation engine, and a planning decision engine. The data fusion engine is used to generate a 3D voxel model of the community, the multi-agent simulation engine is used to drive the evolution of agent behavior, and the planning decision engine is used to perform supply and demand gap analysis of commercial facilities and output optimization schemes.

[0010] The specific working steps of this method are as follows:

[0011] Step 1: Multi-source data fusion and high-precision 3D modeling; On the cloud computing platform, location-based service (LBS) mobile trajectory data, 3D point cloud data collected by UAV oblique photography, multi-period land use attribute data, community population census and real-time update data, social media sentiment analysis data, road network topology vector data, and commercial facility point of interest (POI) data are fused. Through graph theory-based spatial topology consistency verification and redundant data cleaning process, a high-precision 3D voxel model is generated. This model uniformly labels terrain slope values, building outline geometric vectors, and road network topology connections with a spatial resolution of 0.5m × 0.5m. It automatically identifies public passage areas based on a grid accessibility algorithm. Based on the spatial density distribution and directional consistency characteristics of LBS mobile trajectory data, it assists in correcting road network topology connections and identifying high-frequency public passage areas. The overall model's global spatial error is controlled within ±0.15 meters.

[0012] Step Two: Construction of a Multi-Agent Collaborative System and Implantation of Behavioral Rules; Three types of entity models are created: resident intelligent agents, commercial facility intelligent agents, and environmental constraint intelligent agents, and dynamic interaction rules are established among them; the resident intelligent agent is loaded with an age-stratified demand feature matrix based on census data, dividing age groups into children (0-12 years old), adolescents (13-18 years old), young adults (19-40 years old), middle-aged (41-60 years old), and elderly (60 years and above); a decision-making architecture is adopted, including a three-layer mechanism of business type matching screening, spatial accessibility assessment, and dynamic preference weighting: firstly, the matching degree is calculated from the commercial facility point of interest (POI) data based on the preset business type demand matching degree threshold of 0.75, where the decision-making logic for the elderly group is forcibly bound to medical services, as shown in the following formula:

[0013] Score = I base ·P ang ·(1-0.1D category )

[0014] The "Score" represents the matching degree, indicating the priority of a commercial facility in meeting the essential needs of a specific resident intelligent agent. A higher score indicates that the facility's business type better matches the agent's basic needs profile. base P is the benchmark parameter for business type. ang D represents the age-stratified demand preference parameter. category The deviation coefficient between facility and agent demand categories;

[0015] Secondly, spatial accessibility analysis is performed using 3D point cloud data, road network topology vector data, and commercial facility point of interest (POI) data collected through UAV oblique photography, and the effective service radius after incorporating the slope influence coefficient is calculated. Finally, dynamic preference weights generated by social media sentiment analysis data are integrated to output the final facility selection decision.

[0016] The intelligent agent for commercial facilities calculates its overall attractiveness value using a dynamic competition model, specifically employing the following formula:

[0017]

[0018] In the formula A j Let K be the overall attractiveness value of facility j, representing the total attraction intensity of commercial facility j to the resident intelligent agent under a specific simulated spatiotemporal environment; j S is the benchmark coefficient for the business type, where S is the building footprint area and N is the base area of ​​the facility. comp Indicates the number of similar facilities within a 500-meter radius;

[0019] The environmentally constrained agent calculates the three-dimensional impedance of the path to the terrain slope and construction obstacles, providing spatial travel cost parameters for the resident agent, as shown in the following formula:

[0020]

[0021] Where G is the average effective slope impedance value of the path, representing a comprehensive quantitative assessment of the difficulty of pedestrian access for residents due to the terrain undulations of the entire path; n is the total number of path segments; Δh i Let d be the elevation change of the i-th road segment. i This represents the planar distance. A path replanning protocol is triggered when G > 0.8.

[0022] Step 3: The multi-agent spatiotemporal behavior flow simulation engine runs; a high-precision time-progression controller with a basic time step of 15 minutes is configured to drive 100,000 to 150,000 agents to perform concurrent collaborative evolution, simulating 96 consecutive time steps in each simulation day; the system tracks the spatial location and behavioral state of each agent in real time through distributed memory computing, and collects passenger flow data of each commercial facility at 5-second intervals, and calculates the service pressure index in combination with the facility's design capacity; when the service pressure index is obtained by normalizing the ratio of facility passenger flow to design capacity in real time, and when the index is detected to be not less than 1.3 within 3 consecutive simulation steps, the system automatically triggers the service radius contraction protocol, temporarily adjusting the effective service radius of the affected facilities to 0.9 times the original value, and simultaneously generating a facility capacity warning signal;

[0023] Step 4: Diagnosing blind spots in community commercial supply and demand matching; Constructing a quantitative analysis model based on a dual-dimensional demand and supply model, converting community census data and consumption behavior characteristics into commercial service demand equivalents based on the average consumption capacity of standard households; Simultaneously, combining the high-precision three-dimensional voxel model generated in Step 1, calculating the effective supply equivalent of existing Points of Interest (POI) commercial facilities based on spatial accessibility and service carrying capacity; By comparing the numerical differences between demand equivalents and supply equivalents in real time, automatically generating planning and control instructions: When the demand gap rate ΔQ is continuously greater than 0.3, the system marks the geographical unit as a priority area for the layout of new commercial facilities and triggers the planning decision unit to generate alternative site selection schemes;

[0024] Step 5: Multi-objective optimization and flexible scheme generation; A constrained multi-objective evolutionary algorithm and deep reinforcement learning fusion framework are used to generate community commercial facility planning schemes. Based on the supply-demand gap ratio ΔQ output from Step 4, a triple elasticity mechanism of spatial elasticity, business format configuration, and scale expansion is continuously optimized. The agent uses the distribution of the supply-demand gap in the community's geographical space as the state input, and dynamically decides the proportion of reserved land, the intensity of multi-business format combination configuration, and the scale expansion of facilities through a policy neural network. Its core reward function R is designed as a multi-objective trade-off function that balances demand matching accuracy and cost control efficiency, specifically expressed as:

[0025]

[0026] In the formula, R is the core reward function, representing a quantitative score of the overall merits of a planning scheme; e is a natural constant, serving as the base of the exponential function; ω is the demand matching weight coefficient, γ is the cost penalty coefficient, C0 is the benchmark value of construction cost, and η is the stability adjustment parameter for scale changes. This function drives the agent to autonomously explore Pareto optimal planning schemes while ensuring that the coverage rate of the basic needs of the elderly population is not lower than a preset constraint. A spatial elasticity mechanism dynamically reserves commercially compatible construction land based on the land development potential rating. A business format configuration elasticity mechanism directly responds to changes in the spatial gradient of ΔQ and dynamically adjusts the proportion of service formats. A scale expansion elasticity mechanism provides a dynamic adjustment interface to support federated learning collaborative optimization. The final generated planning scheme is empirically verified, with a prediction error of less than 5%, and outputs a three-dimensional spatial decision report integrating all optimization parameters and spatial constraints.

[0027] Furthermore, the multi-source data types and processing specifications input on the cloud computing platform in step one are shown in the following table:

[0028]

[0029]

[0030] Furthermore, the high-precision three-dimensional voxel model generated in step one is a multimodal attribute fusion model based on an octree index structure. Its basic voxel unit is a 0.5m × 0.5m × 0.5m cube. Each voxel unit integrates and stores the terrain slope value, the geometric vector identifier of the outer contour of the building to which it belongs, the topological connection relationship of the road network to which it belongs, and the grid accessibility identifier. The model is jointly adjusted and solved by deploying a high-precision spatial control network, so that its global spatial error is controlled within ±0.15m.

[0031] Furthermore, in step two, the resident agent loads an age-stratified demand feature matrix based on census data, dividing the resident agent into five levels according to age, and predefining the core facility and business type demand preferences for each level. This serves as the basic rule for the business type matching and filtering layer in the agent's decision-making architecture. The specific structure of this matrix is ​​shown in the table below:

[0032]

[0033] Furthermore, the three-layer decision-making mechanism of the resident intelligent agent in step two includes a first-layer screening mechanism of business type matching screening, a second-layer screening mechanism of spatial accessibility assessment, and a third-layer screening mechanism of dynamic preference weighting. The implementation specifications are shown in the table below:

[0034]

[0035]

[0036] Furthermore, in step two, the matching degree is calculated from the commercial facility point of interest (POI) data based on a preset business type demand matching degree threshold of 0.75. The decision-making logic for the elderly group is forcibly linked to medical services, as shown in the following formula:

[0037] Score = I base ·P ang ·(1-0.1D category )

[0038] The "Score" represents the matching degree, indicating the priority of a commercial facility in meeting the essential needs of a specific resident intelligent agent. A higher score indicates that the facility's business type better matches the agent's basic needs profile. base P is the benchmark parameter for business type. ang D represents the age-stratified demand preference parameter. category This refers to the deviation coefficient between facility and intelligent agent demand categories. The value of the business type benchmark parameter was obtained by organizing urban planning and business analysis experts to conduct multiple rounds of Delphi method scoring on the necessity of each business type for people's livelihood based on the national economic industry classification code; the value of the age stratification preference parameter was obtained by using principal component analysis to extract the preference characteristics of each age group in different consumption categories based on the household consumption expenditure structure in the population census data, and then mapping it to a preset numerical range through Min-Max normalization; the value of the demand category deviation coefficient was deterministically assigned by querying a predefined facility business type-demand category mapping rule table, which clearly stipulates the matching relationship between different demand types and facility business types and their corresponding discrete assignments.

[0039] Furthermore, the facility service pressure index P in step three... service The calculation formula is as follows:

[0040] P service =(C current / C capacity ) / N

[0041] Among them, P service Service stress index; C current C represents the real-time visitor flow to a facility within the current simulation step size; capacity The design capacity of the facility is denoted by ; N is the normalization coefficient used to adjust the calculation results to a preset comparable range. The triggering and execution logic of the service radius contraction protocol is as follows: when a facility's P is detected... service When three consecutive time steps are all greater than or equal to 1.3, the service radius shrinkage protocol is automatically triggered, reducing the effective service radius R of the facility. old Updated to R new =0.9×R oldAnd simultaneously generate capacity warning signals.

[0042] Furthermore, the commercial service demand equivalent D in step four is determined as follows: based on community census data, the total number of households N and the average annual consumption expenditure per household C are obtained. D Based on the consumption behavior characteristics exhibited by resident agents in the simulation, the intensity coefficient K of consumption preference for specific business formats for agents of different age groups is extracted. age The formula for calculating the equivalent demand D for the aforementioned commercial services is:

[0043] D = C D ·N·K age ·η

[0044] Wherein, η is an adjustment coefficient determined based on the regional business vitality level, with a value ranging from 0.6 to 1.4; the consumption preference intensity coefficient K age This was obtained by analyzing the fitting of the frequency of business visits and the amount of consumption by the intelligent agent in historical simulation.

[0045] The effective supply equivalent S in step four is determined as follows: Based on the high-precision three-dimensional voxel model generated in step one, the existing commercial facility POI data is quantitatively calculated by integrating its spatial accessibility and service carrying capacity indicators; the formula for calculating the effective supply equivalent S is:

[0046]

[0047] Where A is the service carrying capacity of the facility, which is calculated by A = facility operating area × unit area capacity coefficient × average daily operating hours; μ is the facility business format conversion coefficient, which is determined according to the national economic industry classification code. λ is a spatial accessibility function based on distance decay, where a is the Euclidean distance from the facility to the geometric center of the target geographic unit, and λ is the decay coefficient, which is corrected based on road network complexity and terrain slope data.

[0048] The formula for calculating the demand gap rate ΔQ in step four is: ΔQ=(DS) / D, where D is the demand equivalent of commercial services and S is the effective supply equivalent. When the system detects that the ΔQ value of a certain geographical unit is continuously greater than 0.3, it automatically marks the area as a priority area for the layout of new commercial facilities and triggers the generation of alternative site selection schemes.

[0049] Furthermore, in step five, the deep reinforcement learning fusion framework employs a 7-layer convolutional neural network architecture to process the 3D supply-demand gap tensor input, with a network parameter scale of 1.2 × 10⁻⁶. 6 The input layer dimension of 20×20×0.5m corresponds to the spatial grid accuracy, and the output three-dimensional motion vector [φ,ψ,κ] controls the spatial reservation ratio, business configuration intensity and scale expansion range, respectively.

[0050] Furthermore, the implementation rules of the elastic mechanism in step five are as follows: The spatial elastic mechanism dynamically sets the proportion of reserved land for commercially compatible construction based on the supply-demand gap rate ΔQ. When ΔQ>0.4, 25% of high-intensity land is reserved; when 0.3<ΔQ≤0.4, 20% of medium-intensity land is reserved; and when ΔQ≤0.3, 15% of benchmark land is reserved. The coordinate accuracy of the reserved land boundary is controlled within 0.01 meters and supports visualization projection through AR devices. The business format configuration elastic mechanism implements a gap response strategy, wherein the number of fresh food facilities is determined by an up-rounding function. Dynamically calculate the number of new additions (where ΔN is the formula) k Let ΔQ be the newly added quantity of the k-th type of facility, and ΔQ be the demand gap rate. (Rounding up) and the configuration weight of fresh food facilities in senior communities is increased by an additional 40%, while educational facilities must be matched and configured within 300 meters of schools; the scale expansion elasticity mechanism adopts a modular telescopic architecture, the capacity adjustment range is constrained between 0.3 and 1.5 times the base capacity, and stepless adjustment is achieved through hydraulic support units.

[0051] Beneficial effects of the present invention

[0052] 1. This invention integrates seven types of multi-source data to construct a high-precision three-dimensional voxel model with a global spatial error of less than 0.15 meters, and establishes a multi-agent simulation system that fits the real decision-making logic of residents of different ages. It fundamentally changes the traditional model that relies on static data and experience-based judgment, and enables business demand analysis, blind spot diagnosis and solution generation to be based on dynamic and detailed simulation data, which greatly improves the scientificity and reliability of planning output.

[0053] 2. The system drives the spatiotemporal behavior simulation of 100,000 intelligent agents in 15-minute increments, enabling it to capture real-time changes in service pressure on commercial facilities (such as capacity overload) and automatically diagnose planning blind spots based on accurately calculated supply-demand gap ratios (ΔQ). This dynamic and quantitative diagnostic capability allows planners to promptly identify transient and localized supply-demand imbalances that are difficult to detect through static analysis, triggering the system to automatically generate response plans, thus achieving refined management and forward-looking planning of community commercial vitality.

[0054] 3. This invention employs a framework that integrates deep reinforcement learning and multi-objective evolutionary algorithms to automatically generate planning schemes that simultaneously optimize the three elastic mechanisms of space, business type, and scale. These schemes are validated and driven by quantitative indicators such as commercial service coverage efficiency and facility-population matching. This method overcomes the shortcomings of traditional planning schemes, which are often singular and difficult to adjust. It can quickly output and compare multiple Pareto optimal solutions under multiple constraints, freeing planners from tedious manual analysis and trial and error, significantly improving the efficiency of planning and the overall quality of the schemes. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the method implemented in this invention;

[0056] Figure 2 This is a diagram showing the system module composition of the present invention;

[0057] Figure 3 This is a simulation diagram of the multi-agent system of the present invention;

[0058] Figure 4 This is a spatial planning diagram of community commercial facilities according to the present invention. Detailed Implementation

[0059] 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.

[0060] A method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation, such as Figure 1-4 As shown, it includes the following steps:

[0061] I. Multi-source data fusion and high-precision 3D modeling: On the cloud computing platform, location-based service (LBS) mobile trajectory data, 3D point cloud data collected by UAV oblique photography, multi-period land use attribute data, community population census and real-time update data, social media sentiment analysis data, road network topology vector data, and commercial facility point of interest (POI) data are fused together. Through graph theory-based spatial topology consistency verification and redundant data cleaning processes, a high-precision 3D voxel model is generated. This model uniformly labels terrain slope values, building outline geometric vectors, and road network topology connections with a spatial resolution of 0.5m × 0.5m. It automatically identifies public passage areas based on a raster accessibility algorithm. Based on the spatial density distribution and directional consistency characteristics of LBS mobile trajectory data, it assists in correcting road network topology connections and identifying high-frequency public passage areas. The overall model's global spatial error is controlled within ±0.15 meters.

[0062] The multi-source data types and processing specifications input to the cloud computing platform are shown in the table below:

[0063]

[0064]

[0065] The high-precision three-dimensional voxel model is a multimodal attribute fusion model based on an octree index structure. Its basic voxel unit is a 0.5m × 0.5m × 0.5m cube. Each voxel unit integrates and stores the terrain slope value, the geometric vector identifier of the outer contour of the building to which it belongs, the topological connection relationship of the road network to which it belongs, and the grid accessibility identifier. The model is jointly adjusted by deploying a high-precision spatial control network to control its global spatial error within ±0.15m.

[0066] II. Construction and Behavioral Rule Implantation of a Multi-Agent Collaborative System: Three types of entity models were created: resident intelligent agents, commercial facility intelligent agents, and environmental constraint intelligent agents. Dynamic interaction rules were established among these models. The resident intelligent agent was loaded with an age-stratified demand feature matrix based on census data, dividing age groups into children (0-12 years), adolescents (13-18 years), young adults (19-40 years), middle-aged (41-60 years), and elderly (60 years and above). A decision-making architecture was adopted, incorporating a three-layer mechanism: business type matching screening, spatial accessibility assessment, and dynamic preference weighting. First, the matching degree was calculated from commercial facility points of interest (POI) data based on a preset business type demand matching degree threshold of 0.75. The decision-making logic for the elderly group was forcibly bound to medical services, as shown in the following formula:

[0067] Score = I base ·P ang ·(1-0.1D category )

[0068] The "Score" represents the matching degree, indicating the priority of a commercial facility in meeting the essential needs of a specific resident intelligent agent. A higher score indicates that the facility's business type better matches the intelligent agent's basic needs profile. base P is the benchmark parameter for business type. ang D represents the age-stratified demand preference parameter. category This is the deviation coefficient between facility and agent demand categories.

[0069] Secondly, spatial accessibility analysis is performed by calling up 3D point cloud data, road network topology vector data, and commercial facility point of interest (POI) data collected by UAV oblique photography, and the effective service radius after introducing the slope influence coefficient is calculated. Finally, dynamic preference weights generated by social media sentiment analysis data are integrated to output the final facility selection decision.

[0070] The intelligent agent for commercial facilities calculates its overall attractiveness value using a dynamic competition model, specifically employing the following formula:

[0071]

[0072] In the formula A j Let K be the overall attractiveness value of facility j, representing the total attraction intensity of commercial facility j to resident intelligent agents under a specific simulated spatiotemporal environment. j S is the benchmark coefficient for the business type, where S is the building footprint area and N is the base area of ​​the facility. comp This indicates the number of similar facilities within a 500-meter radius.

[0073] The environmentally constrained agent calculates the three-dimensional impedance of the path to the terrain slope and construction obstacles, providing spatial travel cost parameters for the resident agent, as shown in the following formula:

[0074]

[0075] Where G is the average effective slope impedance value of the path, representing a comprehensive quantitative assessment of the difficulty of pedestrian access for residents due to the terrain undulations of the entire path; n is the total number of path segments; Δh i Let d be the elevation change of the i-th road segment. i This represents the planar distance. A path replanning protocol is triggered when G > 0.8.

[0076] The resident agent is loaded with an age-stratified demand feature matrix based on census data. This matrix divides the resident agent into five age levels and predefines the core facility and business type demand preferences for each level. These preferences serve as the foundational rules for the business type matching and filtering layer in the agent's decision-making architecture. The specific structure of this matrix is ​​shown in the table below:

[0077]

[0078]

[0079] The three-tier decision-making mechanism of the resident intelligent agent includes a first-tier screening mechanism for business type matching, a second-tier spatial accessibility assessment, and a third-tier dynamic preference weighting mechanism. The implementation specifications are shown in the table below:

[0080]

[0081] The matching degree is calculated from commercial facility point of interest (POI) data based on a preset business type demand matching degree threshold of 0.75. The decision-making logic of the elderly group is forcibly linked to medical services, and the formula is as follows:

[0082] Score = I base ·P ang ·(1-0.1D category )

[0083] The "Score" represents the matching degree, indicating the priority of a commercial facility in meeting the essential needs of a specific resident intelligent agent. A higher score indicates that the facility's business type better matches the intelligent agent's basic needs profile. base P is the benchmark parameter for business type. ang D represents the age-stratified demand preference parameter. category This refers to the deviation coefficient between facility and intelligent agent demand categories. The value of the business type benchmark parameter was obtained by organizing urban planning and business analysis experts to conduct multiple rounds of Delphi method scoring on the necessity of each business type for people's livelihood based on the national economic industry classification code; the value of the age stratification preference parameter was obtained by using principal component analysis to extract the preference characteristics of each age group in different consumption categories based on the household consumption expenditure structure in the population census data, and then mapping it to a preset numerical range through Min-Max normalization; the value of the demand category deviation coefficient was deterministically assigned by querying a predefined facility business type-demand category mapping rule table, which clearly stipulates the matching relationship between different demand types and facility business types and their corresponding discrete assignments.

[0084] 3. The multi-agent spatiotemporal behavior flow simulation engine is configured with a high-precision time-progression controller with a basic time step of 15 minutes, driving 100,000 to 150,000 agents to perform concurrent collaborative evolution, simulating 96 consecutive time steps in each simulation day; the system tracks the spatial location and behavioral state of each agent in real time through distributed memory computing, and collects passenger flow data of each commercial facility at 5-second intervals, and calculates the service pressure index in combination with the facility's design capacity; when the service pressure index is obtained by normalizing the ratio of facility passenger flow to design capacity in real time, and the index is detected to be not less than 1.3 within 3 consecutive simulation steps, the system automatically triggers the service radius contraction protocol, temporarily adjusting the effective service radius of the affected facilities to 0.9 times the original value, and simultaneously generating a facility capacity warning signal.

[0085] The facility service pressure index P service The calculation formula is as follows:

[0086] P service =(C current / C capacity ) / N

[0087] Among them, P service Service stress index; C current C represents the real-time visitor flow to a facility within the current simulation step size; capacity The design capacity of the facility is denoted by ; N is the normalization coefficient used to adjust the calculation results to a preset comparable range. The triggering and execution logic of the service radius contraction protocol is as follows: when a facility's P is detected... serviceWhen three consecutive time steps are all greater than or equal to 1.3, the service radius shrinkage protocol is automatically triggered, reducing the effective service radius R of the facility. old Updated to R new =0.9×R old And simultaneously generate capacity warning signals.

[0088] IV. Diagnosing blind spots in community commercial supply and demand matching: Constructing a quantitative analysis model based on a dual-dimensional demand and supply model, converting community census data and consumption behavior characteristics into commercial service demand equivalents based on the average consumption capacity of standard households; Simultaneously, combining the high-precision three-dimensional voxel model generated in step one, calculating the effective supply equivalent of existing Points of Interest (POI) commercial facilities based on spatial accessibility and service carrying capacity; By comparing the numerical differences between demand equivalents and supply equivalents in real time, automatically generating planning and control instructions: When the demand gap rate ΔQ is continuously greater than 0.3, the system marks the geographical unit as a priority area for the layout of new commercial facilities and triggers the planning decision unit to generate alternative site selection schemes.

[0089] The equivalent demand for commercial services, D, is determined by: obtaining the total number of households N and the average annual consumption expenditure per household C based on community census data. D Based on the consumption behavior characteristics exhibited by resident agents in the simulation, the intensity coefficient K of consumption preference for specific business formats for agents of different age groups is extracted. age The formula for calculating the equivalent demand D for the aforementioned commercial services is:

[0090] D = C D ·N·K age ·η

[0091] Wherein, η is an adjustment coefficient determined based on the regional business vitality level, with a value ranging from 0.6 to 1.4; the consumption preference intensity coefficient K age This was obtained by analyzing the fitting of the frequency of business visits and the amount of consumption by the intelligent agent in historical simulation.

[0092] The effective supply equivalent S is determined as follows: Based on the high-precision three-dimensional voxel model generated in step one, the existing commercial facility POI data is quantitatively calculated by integrating its spatial accessibility and service carrying capacity indicators; the formula for calculating the effective supply equivalent S is:

[0093]

[0094] Where A is the service carrying capacity of the facility, which is calculated by A = facility operating area × unit area capacity coefficient × average daily operating hours; μ is the facility business format conversion coefficient, which is determined according to the national economic industry classification code. λ is a spatial accessibility function based on distance decay, where a is the Euclidean distance from the facility to the geometric center of the target geographic unit, and λ is the decay coefficient, which is corrected based on road network complexity and terrain slope data.

[0095] The formula for calculating the demand gap ratio ΔQ is: ΔQ=(DS) / D, where D is the demand equivalent of commercial services and S is the effective supply equivalent. When the system detects that the ΔQ value of a certain geographical unit is continuously greater than 0.3, it automatically marks the area as a priority area for the layout of new commercial facilities and triggers the generation of alternative site selection schemes.

[0096] V. Multi-objective optimization and flexible scheme generation: A constrained multi-objective evolutionary algorithm and deep reinforcement learning fusion framework are used to generate community commercial facility planning schemes. The supply-demand gap ratio ΔQ output from step four continuously optimizes a triple elasticity mechanism encompassing spatial elasticity, business format configuration, and scale expansion. The agent uses the distribution of the supply-demand gap in the community's geographic space as its state input, and dynamically decides on the proportion of reserved land, the intensity of multi-business format combination configuration, and the extent of facility scale expansion through a policy neural network. Its core reward function R is designed as a multi-objective trade-off function that balances demand matching accuracy and cost control efficiency, specifically expressed as:

[0097]

[0098] In the formula, R is the core reward function, representing a quantitative score of the overall merits of a planning scheme; e is a natural constant, serving as the base of the exponential function; ω is the demand matching weight coefficient, γ is the cost penalty coefficient, C0 is the benchmark value of construction cost, and η is the stability adjustment parameter for scale changes. This function drives the agent to autonomously explore the Pareto optimal planning scheme while ensuring that the coverage rate of the basic needs of the elderly population is not lower than a preset constraint. The spatial elasticity mechanism dynamically reserves commercially compatible construction land based on the land development potential rating; the business format configuration elasticity mechanism directly responds to changes in the spatial gradient of ΔQ and dynamically adjusts the proportion of service formats; the scale expansion elasticity mechanism provides a dynamic adjustment interface to support federated learning collaborative optimization. The final generated planning scheme has been empirically verified, with a prediction error of less than 5%, and outputs a three-dimensional spatial decision report integrating all optimization parameters and spatial constraints.

[0099] The deep reinforcement learning fusion framework employs a 7-layer convolutional neural network architecture to process the 3D supply-demand gap tensor input, with a network parameter size of 1.2 × 10⁻⁶. 6 The input layer dimension of 20×20×0.5m corresponds to the spatial grid accuracy, and the output three-dimensional motion vector [φ,ψ,κ] controls the spatial reservation ratio, business configuration intensity and scale expansion range, respectively.

[0100] The implementation rules of the flexible mechanism are as follows: The spatial flexible mechanism dynamically sets the proportion of reserved land for commercially compatible construction based on the supply-demand gap rate ΔQ. When ΔQ>0.4, 25% of high-intensity land is reserved; when 0.3<ΔQ≤0.4, 20% of medium-intensity land is reserved; and when ΔQ≤0.3, 15% of benchmark land is reserved. The coordinate accuracy of the reserved land boundary is controlled within 0.01 meters and supports visualization projection through AR devices. The business format configuration flexible mechanism implements a gap response strategy, wherein the number of fresh food facilities is determined by an up-rounding function. Dynamically calculate the number of new additions (where ΔN is the formula) k Let ΔQ be the newly added quantity of the k-th type of facility, and ΔQ be the demand gap rate. (Rounding up) and the configuration weight of fresh food facilities in senior communities is increased by an additional 40%, while educational facilities must be matched and configured within 300 meters of schools; the scale expansion elasticity mechanism adopts a modular telescopic architecture, the capacity adjustment range is constrained between 0.3 and 1.5 times the base capacity, and stepless adjustment is achieved through hydraulic support units.

[0101] Example

[0102] The technical solution of the present invention will be described in detail using Hexi New Town in Nanjing, Jiangsu Province as an example.

[0103] (1) Taking Hexi New Town in Nanjing, Jiangsu Province as the target area, data on Hexi New Town in Nanjing was acquired, and multi-source data fusion and high-precision 3D modeling were performed, specifically including:

[0104] (1.1) The system user interface device uses the data acquisition unit to acquire and input 3D point cloud data, multi-phase land use attribute data, community population census and real-time update data, social media sentiment analysis data, road network topology vector data, and commercial facility point of interest (POI) data for the Nanjing Hexi New Town area. All data is uploaded to the cloud computing platform through an encrypted transmission channel. The format standards and processing data examples for the seven types of data are shown in the table below:

[0105]

[0106] (1.2) An octree indexed voxel model of Hexi New City with a resolution of 0.5 meters was generated using the data fusion engine of the cloud computing platform. Each voxel unit fused and stored terrain slope, building outline vectors, road topology relationships, and accessibility indicators. Through high-precision spatial control network adjustment calculation, the global spatial error was controlled within ±0.15 meters.

[0107] (2) Constructing a multi-agent collaborative system and implanting behavioral rules for Nanjing Hexi New City, limiting the creation and configuration of resident agents, commercial facility agents, and environmental constraint agents, specifically including:

[0108] (2.1) Generate 120,000 resident intelligent agents. Based on the age-stratified demand characteristic matrix of the Hexi New City population census data, divide the age groups into children (0-12 years old), adolescents (13-18 years old), young adults (19-40 years old), middle-aged (41-60 years old), and elderly (60 years old and above). The specific core facility business format demand preferences and special decision-making rules are shown in the table below:

[0109]

[0110] The resident intelligent agent decision-making architecture adopts a three-layer mechanism including business type matching and screening, spatial accessibility assessment, and dynamic preference weighting. Regarding business type matching and screening, a general matching threshold of 0.75 is set, but the matching threshold for medical facilities specifically set for the elderly is lowered to 0.65 to reduce their access to medical care. (Score = I) base ·P ang ·(1-0.1D category The system calculates the matching degree between business types; if the matching degree between the agent and the facility is lower than this value, it is filtered out. Regarding spatial accessibility assessment, it utilizes 3D point cloud data, road network topology vector data, and commercial facility point of interest (POI) data collected through UAV oblique photography to calculate the effective service radius R = 500 × (1 - 0.15G), filtering out a subset of accessible facilities within the walkable range of Hexi New City. For dynamic preference weighting, it integrates the sentiment polarity values ​​from Weibo and Dianping to generate a weight matrix, employing the Softmax function. The target facility ultimately selected by the agent is determined. The decision-making criteria, processing rules, and output results of the three-layer mechanism are shown in the table below:

[0111]

[0112]

[0113] (2.2) Create 350 intelligent commercial facilities in the Hexi New City area, according to the formula. Dynamically calculate the overall attractiveness, where A j Let K be the overall attractiveness value of facility j, representing the total attraction intensity of commercial facility j to resident intelligent agents under a specific simulated spatiotemporal environment. j S is the benchmark coefficient for the business type, where S is the building footprint area and N is the base area of ​​the facility. comp This indicates the number of similar facilities within a 500-meter radius.

[0114] (2.3) The configuration of environmentally constrained intelligent agents in the Hexi New City area is based on the formula. Calculate the path impedance, where Δh i Let d be the elevation change of the i-th road segment.i This represents the planar distance. A path replanning protocol is triggered when G > 0.8.

[0115] (3) Run the Hexi New City spatiotemporal behavior flow simulation engine in the multi-agent simulation engine of the cloud computing platform, specifically including:

[0116] (3.1) Drive the concurrent evolution of 120,000 agents in Hexi New City for 96 time steps with a step size of 15 minutes. Track the location status of resident agents and commercial facility agents in real time through distributed memory computing.

[0117] (3.2) Collect facility passenger flow data at 5-second intervals, according to formula P service =(C current / C capacity The service stress index is calculated as P / N. service Service stress index; C current C represents the real-time visitor flow to a facility within the current simulation step size; capacity The design capacity of the facility is denoted by ; N is the normalization factor, used to adjust the calculation results to a preset comparable range.

[0118] (3.3) When monitoring a facility in Hexi New City for three consecutive steps P service When the value is ≥1.3, the service radius shrinkage protocol is automatically triggered, temporarily adjusting the effective service radius to 0.9 times the original value, and a capacity warning signal is generated simultaneously.

[0119] (4) Conduct blind spot diagnosis of commercial supply and demand matching in Hexi New Town community within the planning and decision engine of the cloud computing platform, specifically including:

[0120] (4.1) Based on the population census data (statistics based on total number of households) and consumption behavior characteristics of the Hexi New City area, according to the formula D = C D ·N·K age • η is used to calculate the equivalent demand for commercial services, where η is an adjustment coefficient determined based on the regional commercial vitality level, with a value ranging from 0.6 to 1.4; the consumption preference intensity coefficient K age This was obtained by analyzing the fitting of the frequency of business visits and the amount of consumption by the intelligent agent in historical simulation.

[0121] (4.2) Combining the high-precision 3D model of Hexi New City in the 3D display unit of the user interface device, the existing 350 commercial facility intelligent agents are analyzed according to the formula. Calculate the effective supply equivalent, where A is the service carrying capacity of the facility, which is calculated by A = facility operating area × unit area capacity coefficient × average daily operating hours; μ is the facility business format conversion coefficient, which is determined according to the national economic industry classification code. λ is a spatial accessibility function based on distance decay, where a is the Euclidean distance from the facility to the geometric center of the target geographic unit, and λ is the decay coefficient, which is corrected based on road network complexity and terrain slope data.

[0122] (4.3) The gap ratio is calculated using the formula ΔQ = (DS) / D, where D is the demand equivalent for commercial services and S is the effective supply equivalent. When the system detects that a geographic unit ΔQ is consistently greater than 0.3, it is automatically marked as a priority layout area and the generation of alternative site selection schemes is triggered.

[0123] (5) Multi-objective optimization and flexible scheme generation for Hexi New City are performed in the planning and decision engine of the cloud computing platform, specifically including:

[0124] (5.1) Based on the spatial distribution data of the supply-demand gap ratio ΔQ output in step four, the system constructs a three-dimensional supply-demand gap tensor as the state input. This tensor has dimensions of 200×200×1, corresponding to a geographic space with a spatial grid precision of 0.5 meters. A 7-layer convolutional neural network architecture is used for processing, with a network parameter scale of 1.2×10⁻⁶. 6 The output is a three-dimensional action vector [φ,ψ,κ], which controls the proportion of reserved land, the intensity of multi-business combination configuration, and the expansion range of facility scale, respectively. The optimization process is driven by the core reward function R described in claim 1: Wherein, ω is the demand matching weight coefficient (value is 0.7), γ is the cost penalty coefficient (value is 0.2), C0 is the construction cost benchmark (set at 50 million yuan based on local engineering costs), and η is the scale change stability adjustment parameter (value is 0.1).

[0125] (5.2) When ΔQ>0.4, 25% of high-intensity land is reserved; when 0.3<ΔQ≤0.4, 20% of medium-intensity land is reserved; the number of new fresh food facilities is dynamically calculated according to ΔN_k=[10·ΔQ], and the weight of fresh food facility configuration in elderly communities is increased by 40%; the scale expansion range is constrained to 0.3-1.5 times the base capacity.

[0126] (5.3) The final scheme was verified by empirical evidence. Based on the change in the supply-demand gap ratio ΔQ before and after optimization, the formula (ΔQ original scheme - ΔQ new scheme) / ΔQ_ original scheme × 100% was used to quantify that the efficiency of commercial service coverage in Hexi New City was improved by 72%, the matching degree between facilities and target population increased by 64%, the prediction error was <5%, and the Hexi New City three-dimensional spatial decision report with integrated optimization parameters was output.

Claims

1. A method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation, characterized in that: The system employed in this method includes a user interface device, a data transmission network, and a cloud computing platform. The user interface device comprises a data acquisition unit, a 3D display unit, and a human-computer interaction unit. The data acquisition unit is used to acquire and input 3D point cloud data, multi-period land use attribute data, community population census and real-time update data, social media sentiment analysis data, road network topology vector data, and commercial facility point-of-interest data. The 3D display unit is used for visual rendering of planning schemes, and the human-computer interaction unit is used for parameter adjustment. The data transmission network uses an encrypted transmission channel to connect the user interface device and the cloud computing platform. The cloud computing platform includes a data fusion engine, a multi-agent simulation engine, and a planning and decision engine. The data fusion engine is used to generate a three-dimensional voxel model of the community, the multi-agent simulation engine is used to drive the evolution of agent behavior, and the planning and decision engine is used to perform supply and demand gap analysis of commercial facilities and output optimization solutions. Specifically, the steps include the following: Step 1: Multi-source data fusion and high-precision 3D modeling On the cloud computing platform, location-based mobile trajectory data, 3D point cloud data collected by UAV oblique photography, multi-period land use attribute data, community population census and real-time update data, social media sentiment analysis data, road network topology vector data, and commercial facility point of interest data are integrated. Through graph theory-based spatial topology consistency verification and redundant data cleaning processes, a high-precision 3D voxel model is generated. This model uniformly labels terrain slope values, building outline geometric vectors, and road network topology connections with a spatial resolution of 0.5m × 0.5m. It automatically identifies public passage areas based on a grid accessibility algorithm. Based on the spatial density distribution and directional consistency characteristics of LBS mobile trajectory data, it assists in correcting road network topology connections and identifying high-frequency public passage areas. The overall model's global spatial error is controlled within ±0.15 meters. Step Two: Construction of Multi-Agent Collaborative System and Implantation of Behavioral Rules Three types of entity models are created: resident intelligent agents, commercial facility intelligent agents, and environmental constraint intelligent agents, and dynamic interaction rules are established among them. The resident intelligent agent is loaded with an age-stratified demand feature matrix based on census data, dividing age groups into children, adolescents, young adults, middle-aged adults, and the elderly. A decision-making architecture is adopted, including a three-layer mechanism of business type matching screening, spatial accessibility assessment, and dynamic preference weighting: First, the matching degree is calculated from commercial facility interest data based on a preset business type demand matching degree threshold of 0.

75. The decision-making logic for the elderly group is forcibly bound to medical services, as shown in the following formula: Score=I base ·P ang ·(1-0.1D category ) The "Score" represents the matching degree, indicating the priority of a commercial facility in meeting the essential needs of a specific resident intelligent agent. A higher score indicates that the facility's business type better matches the agent's basic needs profile. base P is the benchmark parameter for business type. ang D represents the age-stratified demand preference parameter. category The deviation coefficient between facility and agent demand categories; Secondly, spatial accessibility analysis is performed by calling up 3D point cloud data, road network topology vector data, and commercial facility point of interest data collected by UAV oblique photography, and the effective service radius after introducing the slope influence coefficient is calculated. Finally, dynamic preference weights generated by social media sentiment analysis data are integrated to output the final facility selection decision. The intelligent agent for commercial facilities calculates its overall attractiveness value using a dynamic competition model, specifically employing the following formula: In the formula A j Let K be the overall attractiveness value of facility j, representing the total attraction intensity of commercial facility j to the resident intelligent agent under a specific simulated spatiotemporal environment; j S is the benchmark coefficient for the business type, where S is the building footprint area and N is the base area of ​​the facility. comp Indicates the number of similar facilities within a 500-meter radius; The environmentally constrained agent calculates the three-dimensional impedance of the path to the terrain slope and construction obstacles, providing spatial travel cost parameters for the resident agent, as shown in the following formula: Where G is the average effective slope impedance value of the path, representing a comprehensive quantitative assessment of the difficulty of pedestrian access for residents due to the terrain undulations of the entire path; n is the total number of path segments; Δh i Let d be the elevation change of the i-th road segment. i This represents the planar distance; a path replanning protocol is triggered when G > 0.

8. Step 3: Running the Multi-Agent Spatiotemporal Behavior Flow Simulation Engine The system is equipped with a high-precision time-progression controller with a basic time step of 15 minutes, driving 100,000 to 150,000 agents to perform concurrent and collaborative evolution. Each simulation day fully simulates 96 consecutive time steps. The system tracks the spatial location and behavioral state of each agent in real time through distributed memory computing, and collects passenger flow data of each commercial facility at 5-second intervals. Combined with the facility design capacity, the system calculates the service pressure index. When the service pressure index is obtained by normalizing the ratio of facility passenger flow to design capacity in real time, and the index is detected to be no less than 1.3 within 3 consecutive simulation steps, the system automatically triggers the service radius contraction protocol, temporarily adjusts the effective service radius of the affected facilities to 0.9 times the original value, and generates a facility capacity warning signal at the same time. Step 4: Diagnosing blind spots in the supply and demand matching of community commerce A quantitative analysis model based on demand and supply is constructed to convert community census data and consumption behavior characteristics into commercial service demand equivalents based on the average consumption capacity of standard households. At the same time, combined with the high-precision three-dimensional voxel model generated in step one, the effective supply equivalent of existing point-of-interest commercial facilities based on spatial accessibility and service carrying capacity is calculated. By comparing the numerical differences between demand equivalents and supply equivalents in real time, planning and control instructions are automatically generated: when the demand gap rate ΔQ is continuously greater than 0.3, the system marks the geographical unit as a priority area for the layout of new commercial facilities and triggers the planning decision unit to generate alternative site selection schemes. Step 5: Multi-objective optimization and flexible solution generation A constrained multi-objective evolutionary algorithm and deep reinforcement learning framework are used to generate community commercial facility planning schemes. The scheme continuously optimizes spatial elasticity, business format configuration, and scale expansion based on the supply-demand gap ratio ΔQ output from step four. The agent uses the distribution of the supply-demand gap in the community's geographic space as its state input and dynamically decides on the proportion of reserved land, the intensity of multi-business format configuration, and the extent of facility scale expansion through a policy neural network. Its core reward function R is designed as a multi-objective trade-off function that balances demand matching accuracy and cost control efficiency, specifically expressed as: In the formula, R is the core reward function, which represents a quantitative score of the overall merits of a planning scheme; e is a natural constant, serving as the base of the exponential function; ω is the demand matching weight coefficient, γ is the cost penalty coefficient, C0 is the benchmark value of construction cost, and η is the stability adjustment parameter for scale changes. This function drives the agent to autonomously explore Pareto optimal planning schemes while ensuring that the coverage rate of the basic needs of the elderly population is not lower than a preset constraint. A spatial elasticity mechanism dynamically reserves commercially compatible construction land based on the development potential rating of land parcels. A business format configuration elasticity mechanism directly responds to changes in the spatial gradient of ΔQ and dynamically adjusts the proportion of service formats. A scale expansion elasticity mechanism provides a dynamic adjustment interface to support federated learning collaborative optimization. The final generated planning scheme has been empirically verified, with a prediction error of less than 5%, and outputs a three-dimensional spatial decision report integrating all optimization parameters and spatial constraints.

2. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 1, characterized in that, The multi-source data types and processing specifications input on the cloud computing platform in step one are shown in the table below:

3. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 2, characterized in that, The high-precision three-dimensional voxel model generated in step one is a multimodal attribute fusion model based on an octree index structure. Its basic voxel unit is a 0.5m × 0.5m × 0.5m cube. Each voxel unit integrates and stores the terrain slope value, the geometric vector identifier of the outer contour of the building to which it belongs, the topological connection relationship of the road network to which it belongs, and the grid accessibility identifier. The model is jointly adjusted by deploying a high-precision spatial control network to control its global spatial error within ±0.15m.

4. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 3, characterized in that, Step two involves loading an age-stratified demand feature matrix based on census data into the resident agent. This matrix divides the resident agent into five age levels and predefines the core facility and business type demand preferences for each level. This serves as the basic rule for the business type matching and filtering layer in the agent's decision-making architecture. The specific structure of this matrix is ​​shown in the table below.

5. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 4, characterized in that, The three-tiered decision-making mechanism of the resident intelligent agent in step two includes a first-tier screening mechanism of business type matching screening, a second-tier spatial accessibility assessment mechanism, and a third-tier dynamic preference weighting mechanism. The implementation specifications are shown in the table below:

6. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 5, characterized in that, In step two, the matching degree is calculated from the commercial facility interest point data based on the preset business demand matching degree threshold of 0.

75. The decision-making logic of the elderly group is forcibly linked to medical services, and the formula is as follows: Score=I base ·P ang ·(1-0.1D category ) The "Score" represents the matching degree, indicating the priority of a commercial facility in meeting the essential needs of a specific resident intelligent agent. A higher score indicates that the facility's business type better matches the agent's basic needs profile. base P is the benchmark parameter for business type. ang D represents the age-stratified demand preference parameter. category The deviation coefficient for the demand categories of facilities and intelligent agents; the values ​​of the benchmark parameters for business types are obtained by organizing urban planning and business analysis experts to conduct multiple rounds of Delphi method scoring on the degree of necessity for people's livelihood of each business type based on the national economic industry classification code; The values ​​of the age-stratified preference parameters are based on the household consumption expenditure structure in the census data. Principal component analysis is used to extract the preference characteristics of each age group in different consumption categories, and then the values ​​are mapped to a preset numerical range through Min-Max normalization. The value of the demand category deviation coefficient is deterministically assigned by querying a predefined facility type-demand category mapping rule table. This rule table clearly defines the matching relationship between different demand types and facility types and their corresponding discrete assignments.

7. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 6, characterized in that, Step 3, Facility Service Pressure Index P service The calculation formula is as follows: P service =(C current / C capacity ) / N Among them, P service Service stress index; C current C represents the real-time visitor flow to a facility within the current simulation step size; capacity The design capacity of the facility is denoted by ; N is the normalization coefficient used to adjust the calculation results to a preset comparable range; the triggering and execution logic of the service radius contraction protocol is as follows: when a facility's P is detected... service When three consecutive time steps are all greater than or equal to 1.3, the service radius shrinkage protocol is automatically triggered, reducing the effective service radius R of the facility. old Updated to R new =0.9×R old And simultaneously generate capacity warning signals.

8. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 7, characterized in that, The commercial service demand equivalent D in step four is determined by the following method: obtaining the total number of households N and the average annual consumption expenditure per household C based on community census data. D Based on the consumption behavior characteristics exhibited by resident agents in the simulation, the intensity coefficient K of consumption preference for specific business formats for agents of different age groups is extracted. age The formula for calculating the equivalent demand D for the aforementioned commercial services is: D=C D ·N·K age ·or Wherein, η is an adjustment coefficient determined based on the regional business vitality level, with a value ranging from 0.6 to 1.4; the consumption preference intensity coefficient K age This was obtained by analyzing the fitting relationship between the frequency of business visits and the amount of consumption by the intelligent agent in historical simulations; The effective supply equivalent S in step four is determined as follows: Based on the high-precision three-dimensional voxel model generated in step one, the existing commercial facility POI data is quantitatively calculated by integrating its spatial accessibility and service carrying capacity indicators; the formula for calculating the effective supply equivalent S is: Where A is the service carrying capacity of the facility, which is calculated by A = facility operating area × unit area capacity coefficient × average daily operating hours; μ is the facility business format conversion coefficient, which is determined according to the national economic industry classification code. λ is a spatial accessibility function based on distance decay, where a is the Euclidean distance from the facility to the geometric center of the target geographic unit, and λ is the decay coefficient, which is corrected according to the road network complexity and terrain slope data. The formula for calculating the demand gap rate ΔQ in step four is: ΔQ=(DS) / D, where D is the demand equivalent of commercial services and S is the effective supply equivalent. When the system detects that the ΔQ value of a certain geographical unit is continuously greater than 0.3, it automatically marks the area as a priority area for the layout of new commercial facilities and triggers the generation of alternative site selection schemes.

9. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 8, characterized in that, In step five, the deep reinforcement learning fusion framework employs a 7-layer convolutional neural network architecture to process the 3D supply-demand gap tensor input, with a network parameter size of 1.2 × 10⁻⁶. 6 The input layer dimension of 20×20×0.5m corresponds to the spatial grid accuracy, and the output three-dimensional motion vector [φ,ψ,κ] controls the spatial reservation ratio, business configuration intensity and scale expansion range, respectively.

10. The method for spatial planning and layout of community commercial facilities based on urban multi-agent simulation according to claim 9, characterized in that, The implementation rules for the flexible mechanism in step five are as follows: The spatial flexible mechanism dynamically sets the proportion of reserved land for commercially compatible construction based on the supply-demand gap rate ΔQ. When ΔQ>0.4, 25% of high-intensity land is reserved; when 0.3<ΔQ≤0.4, 20% of medium-intensity land is reserved; and when ΔQ≤0.3, 15% of benchmark land is reserved. The coordinate accuracy of the reserved land boundary is controlled within 0.01 meters and supports visualization projection via AR devices. The business configuration flexible mechanism implements a gap response strategy, wherein the number of fresh food facilities is determined by an up-rounding function. The number of new additions is calculated dynamically, where ΔN is the formula. k Let ΔQ be the newly added quantity of the k-th type of facility, and ΔQ be the demand gap rate. The system uses an up-rounding notation and increases the weight of fresh food facilities by an additional 40% in senior living communities. Educational facilities must be configured within 300 meters of schools. The scale expansion elasticity mechanism adopts a modular telescopic architecture, with the capacity adjustment range constrained between 0.3 and 1.5 times the base capacity, and stepless adjustment is achieved through hydraulic support units.

Images