Urban low-altitude multi-scene take-off and landing facility collaborative layout space-time constraint analysis system

Abstract

The application discloses a kind of urban low-altitude multi-scene take-off and landing facility coordinated layout space-time constraint analysis system, it is related to urban low-altitude traffic infrastructure management technical field, including cloud management center, the cloud management center communication connection has following module: scene coordination perception module, integrated multi-agent reinforcement learning and deep Q network, simulate the space-time demand dynamic of four types of scenes of target area, generate scene coordination degree matrix and peak-shaving scheduling strategy, identify the time period and area of sharable facility and output sharing candidate set.The application can dynamically simulate and identify the demand difference and complementarity of four types of scenes, emergency rescue, logistics distribution, tourism and government inspection in space-time dimension by integrating multi-agent reinforcement learning and deep Q network technology, by constructing scene coordination degree matrix and peak-shaving scheduling strategy, the time period and area of sharable facility can be accurately identified, so as to avoid facility duplication and time conflict.

Description

Technical Field

[0001] This invention relates to the field of urban low-altitude transportation infrastructure management technology, specifically to a spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities. Background Technology

[0002] With the increasing application of low-altitude aircraft in emergency rescue, logistics distribution, tourism, and government inspection, the scientific layout of take-off and landing facilities has become a key issue in improving operational efficiency and ensuring safety and controllability. Different application scenarios differ in terms of time distribution, spatial coverage, and service requirements. At the same time, they are subject to the cross-influence of multiple constraints such as environmental noise, urban wind fields, airspace control, and ground conditions. Traditional single-scenario, static planning methods are difficult to adapt to the actual needs of multi-task collaboration and dynamic adaptation.

[0003] For example, the low-altitude take-off and landing facility planning method based on maximum coverage and minimum resistance proposed in Chinese Patent Publication No. CN120047015A establishes a coverage-efficiency dual-objective optimization algorithm based on spatiotemporal coupling. Under the premise of ensuring basic services, it realizes dynamic matching between facility spatial layout and traffic demand, and obtains low-altitude take-off and landing facility planning results.

[0004] Existing technologies for planning urban low-altitude take-off and landing facilities often employ static, single-scenario site selection models. These models fail to fully consider the dynamic differences in demand and collaborative potential across various application scenarios, such as emergency rescue, logistics distribution, tourism, and government inspection, in both time and space dimensions. This results in low matching between facility layout and actual composite needs, insufficient resource utilization, and potential time-based conflicts and functional interference between different scenarios. Furthermore, existing planning methods often neglect the coupled effects of multi-dimensional spatiotemporal constraints in urban environments, such as noise propagation, complex wind fields, airspace restrictions, and ground bearing capacity. Consequently, the site selection results face challenges in feasibility, safety risks, and social acceptance during implementation. It is difficult to achieve multi-scenario infrastructure sharing and optimal comprehensive benefits while ensuring operational safety and efficiency. Therefore, this paper proposes a spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities to address the aforementioned problems. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: a spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities, including a cloud management center, wherein the cloud management center is communicatively connected to the following modules:

[0006] The scene collaboration perception module comprehensively utilizes multi-agent reinforcement learning and deep Q-network to simulate the spatiotemporal demand dynamics of four types of scenarios in the target area, generate a scene collaboration degree matrix and a staggered scheduling strategy, identify the time period and area of ​​shareable facilities and output a shared candidate set, providing multi-scenario fusion demand input for subsequent deployment;

[0007] The constraint fusion assessment module integrates physical information neural networks and multidimensional constraint analysis technology. Based on a shared candidate set, it predicts wind field and noise propagation in complex urban environments. Combining airspace and land use conditions, it constructs an intensity distribution map containing multidimensional constraints, providing high-precision constraint assessment for site selection.

[0008] The multi-objective optimization modeling module is used to combine the scenario synergy matrix, shared candidate set and intensity distribution map to build a multi-objective optimization model with the objectives of maximizing service coverage, minimizing system cost and minimizing negative externalities. It incorporates scenario synergy and constraint data to achieve facility layout optimization under multi-scenario sharing, thereby improving the overall efficiency and fairness of the system.

[0009] The dynamic iterative solution module uses the particle swarm optimization algorithm to dynamically iterate and optimize the multi-objective optimization model. It combines real-time constraint updates and demand fluctuations to adjust the layout scheme. It provides a non-dominated solution set through Pareto front analysis, supports multi-scheme comparison and dynamic adjustment, enhances the system's adaptability and foresight, and outputs the optimal optimization scheme that takes into account safety, efficiency and benefits, thereby improving the system's adaptability, foresight and decision-making scientificity.

[0010] The scheme simulation and verification module conducts simulations of optimization schemes in the target area, evaluates the facility operation efficiency, safety and economy under different scenarios, and proposes targeted suggestions and implementation paths from the dimensions of planning, land use and operation based on the simulation results, forming an operable scheme toolkit.

[0011] Preferably, the scene collaborative perception module includes an intelligent agent interaction simulation unit and a collaborative degree matrix generation unit;

[0012] The intelligent agent interaction simulation unit is used to introduce a multi-agent reinforcement learning framework into the target area, construct an intelligent agent model covering four scenarios: emergency rescue, logistics distribution, tourism and government inspection, simulate the demand interaction and fluctuation patterns in the spatiotemporal dimension, identify demand hotspots and conflict periods, realize dynamic demand mapping, improve the collaborative perception capability between scenarios, and avoid redundant construction of facilities and time period conflicts.

[0013] The synergy matrix generation unit learns historical demand data for each scenario based on a deep Q-network, analyzes demand fluctuation patterns and identifies spatiotemporal complementarity between scenarios, generates a scenario synergy matrix and a peak-shaving scheduling strategy, identifies time periods and regions where facilities can be shared among multiple scenarios, and outputs the identified shared candidate set, providing a quantitative basis for the sharing of facilities across multiple scenarios and improving resource utilization efficiency.

[0014] Preferably, the execution steps of the intelligent agent interaction simulation unit are as follows:

[0015] Based on the city's overall plan or pilot areas for low-altitude economic development, the target area is determined, and historical spatiotemporal data covering four scenarios—emergency rescue, logistics and distribution, tourism and sightseeing, and government inspection—are obtained to ensure that the demand analysis and planning boundaries are accurately matched, providing a reliable data foundation for subsequent modeling.

[0016] Construct intelligent agent models for four scenarios in the target area. Each intelligent agent model models demand behavior based on historical spatiotemporal data, simulates its task triggering logic and movement rules in the target area, realizes the dynamic and realistic expression of demand in various scenarios, and provides core model support for identifying conflict and collaboration potential.

[0017] A multi-agent reinforcement learning framework is introduced, and interaction rules and resource competition mechanisms between agent models are set. Through time-series simulation, the demand conflicts and overlaps in different scenarios during peak hours and hot spots are identified, revealing the actual spatiotemporal competition relationship of multiple scenarios, quantifying the intensity of conflicts, and providing a direct basis for resource staggering and sharing.

[0018] Based on simulation results, a heat map of the spatiotemporal distribution of scenario requirements and a matrix of conflict periods are dynamically generated. A set of facility requirements that can be merged or used in staggered periods is extracted, providing a dynamic requirement mapping basis for multi-scenario collaborative layout. The simulation results are transformed into structured requirement inputs that are visible, quantifiable, and can be directly used to optimize the layout, thereby improving the targeting of planning.

[0019] Preferably, the execution steps of the synergy matrix generation unit are as follows:

[0020] Historical demand sequence data for each scenario is extracted, and a deep Q-network is used to extract time-series features and recognize patterns in the historical demand sequences of each scenario. The fluctuation cycle, intensity changes and spatial clustering patterns of demand are analyzed, and the spatiotemporal complementarity between scenarios is quantified. Through the extraction of time-series features, the periodic fluctuation patterns of demand in each scenario can be accurately identified, providing a reliable foundation for subsequent complementarity analysis.

[0021] Based on the spatiotemporal complementarity analysis results, a multi-scenario synergy matrix is ​​constructed. The matrix elements represent the degree of sharing between any two scenarios in terms of time period and region. Based on this, a staggered scheduling strategy and time period allocation scheme are generated. The constructed synergy matrix can intuitively reflect the spatiotemporal sharing potential between scenarios and generate an operational staggered scheduling scheme to improve facility synergy efficiency.

[0022] Based on the synergy matrix and the off-peak scheduling strategy, candidate areas and time slot combinations that meet the conditions for sharing facilities in multiple scenarios are identified, and a structured shared candidate set is output. The output shared candidate set provides clear candidate area and time slot combinations for subsequent site selection and facility configuration, effectively supporting collaborative layout decisions in multiple scenarios.

[0023] Preferably, the constraint fusion evaluation module includes a wind noise environment prediction unit and a multi-dimensional constraint overlay unit;

[0024] The wind noise environment prediction unit, based on the fusion of physical information neural network with fluid dynamics equations and meteorological data, combined with the time period and regional information of shareable facilities in the shared candidate set, simulates the urban wind field distribution and noise propagation path, predicts the wind field distribution and noise propagation under the urban terrain, realizes dynamic analysis of wind and noise environment risks, improves site selection safety and environmental adaptability, accurately identifies wind noise risk areas, and significantly improves site selection safety and environmental compliance.

[0025] The multi-dimensional constraint overlay unit is used to integrate multi-dimensional constraints including wind, noise, airspace, and land use. By overlaying the constraints in GIS space, it generates an intensity distribution map of the multi-dimensional constraints, forming a visualized constraint map to reduce the risk of implementation and achieve visualized analysis of constraints, thereby effectively reducing the risk of implementing the planning scheme.

[0026] Preferably, the execution steps of the wind noise environment prediction unit are as follows:

[0027] A city wind field prediction model is constructed based on physical information neural network. The Navier-Stokes equation and high-resolution meteorological data are integrated to simulate the spatiotemporal distribution of wind speed, wind direction and turbulence intensity under complex terrain, improve the prediction accuracy of low-altitude wind field, and effectively support the safe site selection of take-off and landing points.

[0028] By combining the aircraft noise source model and acoustic propagation equation, a noise impact range prediction submodule is established. Based on the takeoff and landing frequency and aircraft parameters, the noise contour distribution is dynamically calculated to achieve a quantitative assessment of noise impact and assist in the formulation of noise reduction and avoidance strategies.

[0029] By spatiotemporally overlaying wind field simulation results with noise propagation predictions, a joint risk map is generated, identifying high-risk areas of wind shear and overlapping areas of noise sensitivity. This enables a two-dimensional dynamic assessment of environmental risks, identifies high-comprehensive-risk areas, and provides environmental constraints for layout optimization.

[0030] Preferably, the execution steps of the multidimensional constraint overlay unit are as follows:

[0031] High-risk wind shear areas and noise-sensitive overlapping areas are extracted and integrated with airspace control layers, land use data and building load information to construct a standardized multidimensional constraint dataset, realize unified storage of multi-source constraint data, and provide structurally consistent input for subsequent weighted fusion;

[0032] Using GIS spatial overlay analysis, weighted fusion and conflict detection are performed on various constraint layers to generate a multi-dimensional constraint intensity distribution map, which intuitively reflects the suitability and restriction level of construction in different areas.

[0033] Spatial matching of the constraint intensity distribution map with the shared candidate set is performed to eliminate areas with excessive constraints, output a suitability score matrix, and screen out feasible candidate areas that meet safety and environmental requirements, providing a quantitative basis for site selection optimization.

[0034] Preferably, the execution steps of the multi-objective optimization modeling module are as follows:

[0035] With the goals of maximizing service coverage, minimizing system cost, and minimizing negative externalities, a multi-objective optimization model is constructed. Service coverage is measured by demand satisfaction rate and average response time. System cost includes the economic cost of the entire life cycle. Negative externalities consider noise impact and security risks. The multi-objective optimization model with three major objectives is constructed to achieve a three-dimensional balance decision of service efficiency, economic benefits and security impact.

[0036] Using the scenario synergy matrix, shared candidate set, and multidimensional constraint strength distribution map as model inputs, the decision variables are defined as facility location, type, service period, and shared allocation scheme. By combining synergy, candidate set, and constraint strength map, the decision variables are quantified to provide structured input for dynamic layout optimization.

[0037] Set model constraints, including demand coverage threshold, maximum construction cost, noise safety limit and airspace availability, to ensure that the solution meets actual operation and safety requirements. By embedding hard constraints, ensure that the solution is practically operable and balances safety compliance with cost control.

[0038] Preferably, the execution steps of the dynamic iterative solution module are as follows:

[0039] The particle swarm optimization algorithm is used to solve the multi-objective optimization model. The constraint handling mechanism and dynamic inertia weight strategy are embedded to support efficient optimization in complex solution space, significantly improve the convergence speed of the algorithm, and stably approach the Pareto front within 500 generations.

[0040] During the iteration process, the updated demand data and constraint information are accessed in real time, the facility layout scheme is dynamically adjusted, and the non-dominated solution set is selected by Pareto sorting to realize the dynamic adaptation of the layout scheme to the environment, ensuring the real-time feasibility and robustness of the solution set.

[0041] The Pareto front solution set is analyzed for diversity preservation and convergence, and several optimal optimization schemes that take into account safety, efficiency and effectiveness are output. Multiple sets of equilibrium optimization schemes are provided to support decision-makers to flexibly choose the implementation path according to actual preferences.

[0042] Preferably, the execution steps of the scheme deduction and verification module are as follows:

[0043] Based on the optimal optimization scheme of the output, a discrete event simulation model is constructed in the target area to simulate the dynamic scheduling and queuing process of multiple scenarios in the facility network, so as to realize the fine reproduction of the scheme operation process and expose potential conflicts and bottlenecks in advance.

[0044] Multiple operational scenarios were set up to simulate and evaluate verification indicators including facility utilization, average response time, noise impact range, wind shear accident probability, and full-cycle economic cost, and to quantify the robustness and economic feasibility of the scheme under various future conditions.

[0045] Based on the simulation results, system bottlenecks and potential risks are identified. Implementation suggestions and dynamic adjustment strategies are generated from three dimensions: planning integration, land use coordination, and operation management. This forms a deliverable optimization solution toolkit, providing implementable adjustment measures and control rules to improve the actual operability of the solution.

[0046] This invention provides a spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities. It has the following beneficial effects:

[0047] (i) The spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities, by integrating multi-agent reinforcement learning and deep Q-network technology, can dynamically simulate and identify the differences and complementarities in the spatiotemporal dimensions of four types of scenarios: emergency rescue, logistics distribution, tourism and sightseeing and government inspection. By constructing a scenario coordination degree matrix and a staggered scheduling strategy, it can accurately identify the time periods and areas of shared facilities, thereby avoiding redundant construction of facilities and time period conflicts.

[0048] (II) This urban low-altitude multi-scenario take-off and landing facility collaborative layout spatiotemporal constraint analysis system introduces physical information neural network and GIS spatial analysis technology to quantitatively evaluate multi-dimensional constraints such as environment, airspace, and physics. Through accurate simulation of complex urban wind fields and noise propagation paths, combined with constraints such as airspace control, land use, and building load-bearing capacity, it generates multi-dimensional constraint intensity distribution maps and risk level zoning. It can identify high-risk areas in the early planning stage and propose corresponding protection or restriction suggestions, significantly reducing safety risks and social acceptance barriers in the implementation of facilities.

[0049] (III) This urban low-altitude multi-scenario take-off and landing facility collaborative layout spatiotemporal constraint analysis system constructs a full life cycle cost-benefit analysis framework covering the construction and operation periods. Combined with a multi-objective optimization model, it minimizes the total system cost while meeting service coverage and safety constraints. Through facility sharing and time-shifting strategies, it can effectively improve facility utilization and reduce redundant investment, thereby achieving a comprehensive balance between economic and social benefits under a limited budget, and improving the overall investment return rate and sustainable operation capability of low-altitude infrastructure.

[0050] (iv) The spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities adopts a multi-objective particle swarm optimization algorithm and a real-time data update mechanism. It can seek the Pareto optimal solution set among the three objectives of maximizing service coverage, minimizing system cost and minimizing negative externalities. Through dynamic iteration and frontier solution analysis, it provides a variety of non-dominated layout schemes and supports scheme comparison and parameter sensitivity analysis, thereby enhancing the scientific nature and flexibility of planning decisions and enabling the final scheme to achieve the optimal balance between safety, efficiency and benefits. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the workflow of a spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to the present invention.

[0052] Figure 2 This is a data flow diagram of a spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1, please refer to Figure 1 , Figure 2 This invention provides a technical solution: a spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities, including a cloud management center, which has the following communication connections:

[0055] The scene collaborative perception module comprehensively utilizes multi-agent reinforcement learning and deep Q-network to simulate the spatiotemporal demand dynamics of four types of scenarios in the target area, generate a scene collaboration degree matrix and a peak scheduling strategy, identify the time period and area of ​​shareable facilities and output a shared candidate set, providing multi-scenario fusion demand input for subsequent deployment. The scene collaborative perception module includes an agent interaction simulation unit and a collaboration degree matrix generation unit.

[0056] The intelligent agent interaction simulation unit is used to introduce a multi-agent reinforcement learning framework into the target area, constructing intelligent agent models covering four scenarios: emergency rescue, logistics distribution, tourism, and government inspection. It simulates the demand interaction and fluctuation patterns in the spatiotemporal dimensions, identifies demand hotspots and conflict periods, achieves dynamic demand mapping, enhances the collaborative perception capability between scenarios, avoids redundant facility construction and time period conflicts, significantly reduces the rate of redundant facility construction, and improves the timeliness of demand response and the efficiency of collaborative scheduling. Based on the city's overall plan or low-altitude economic development pilot area, the target area is determined, and historical spatiotemporal data covering the four scenarios (emergency rescue, logistics distribution, tourism, and government inspection) is acquired to ensure accurate matching between demand analysis and planning boundaries, providing a reliable data foundation for subsequent modeling. Intelligent agent models for the four scenarios in the target area are constructed, with each intelligent agent model based on... Historical spatiotemporal data is used to model demand behavior, simulating the task triggering logic and movement patterns within the target area. This enables a realistic representation of dynamic demands across various scenarios, providing core model support for identifying conflict and collaborative potential. A multi-agent reinforcement learning framework is introduced, setting interaction rules and resource competition mechanisms among agent models. Through time-series simulation, demand conflicts and overlaps in different scenarios during peak hours and hotspot areas are identified, revealing the actual spatiotemporal competitive relationships of multiple scenarios, quantifying conflict intensity, and providing direct evidence for resource staggering and sharing. Based on simulation results, a spatiotemporal distribution heatmap of scenario demands and a conflict period matrix are dynamically generated. A set of facility demands that can be merged or staggered is extracted, providing a dynamic demand mapping basis for multi-scenario collaborative layout. The simulation results are transformed into visible, quantifiable, and directly applicable structured demand inputs for optimizing layout, improving the targeting of planning.

[0057] The specific work involves: based on urban master planning documents or low-altitude economic development pilot zone zoning, accurately defining the geographical boundaries of the target area; using a high-precision administrative division vector layer (GeoJSON format) as a base, acquiring historical spatiotemporal data for four scenarios within the area: emergency rescue, logistics and distribution, tourism, and government inspection. The data accuracy requirements are: time granularity no less than the hour level, and spatial accuracy no less than a 100-meter grid. Specifically, the historical spatiotemporal data for the emergency rescue scenario includes emergency event data, resource distribution data, and population and social data; the historical spatiotemporal data for the logistics and distribution scenario includes order and flow data, network and facility data, and traffic and demand data; and the historical spatiotemporal data for the tourism scenario... The data includes tourist behavior data, attraction and facility data, and activity and event data. Historical spatiotemporal data for government inspection scenarios includes inspection task data, urban component data, and event reporting data. Emergency event data covers the time, location, type, level, response time, and deployed resources of historical emergencies (such as medical emergency, fire, and accident rescue). Resource distribution data covers the coordinates, service capacity, and duty status of key emergency facilities such as hospitals, fire stations, and emergency command centers. Population and social data covers population density distribution, distribution of vulnerable groups (such as the elderly), and locations of key protection units (schools, shopping malls). Order and flow data covers the origin and destination of orders and delivery volume for major logistics companies (such as SF Express and JD.com). The data includes time requirements, cargo types, and network and facility data covering the location, capacity, and service area of ​​distribution centers, delivery stations, and express lockers; traffic and demand data covering peak traffic periods, peak order periods during e-commerce promotions, and distribution of delivery demand for special commodities (such as fresh produce and medicine); tourist behavior data covering the time distribution of visitor flow in scenic areas (hourly / daily / seasonal), tourist origins, tour routes, and length of stay; attraction and facility data covering the coordinates, opening hours, carrying capacity, and ticket prices of A-level scenic areas, viewing platforms, and tourist distribution centers; activity and event data covering the time, location, and scale of participation in large-scale festivals, competitions, and exhibitions; and inspection task data covering the inspection routes, frequency, and key points of departments such as urban management, environmental protection, and water conservancy. The inspection targets (such as illegal buildings, sewage outlets, and waterways) and urban component data cover the spatial distribution and status information of municipal facilities (streetlights, manhole covers), public spaces, and construction sites. The event reporting data covers the type, location, and processing time of events reported by citizen hotlines and grid workers. Based on the integrated historical spatiotemporal data, intelligent agent models are constructed for four types of scenarios. Each intelligent agent model defines its behavioral logic through parameterized configuration. A multi-agent reinforcement learning framework is introduced, a unified analog clock step size (5 minutes) is set, and the interaction rules between intelligent agents are defined. The intelligent agents update their strategies by interacting with the environment and other intelligent agents to maximize their own task completion efficiency.Within the constructed reinforcement learning framework, a time-series simulation was conducted using the target region as a virtual sandbox for a typical period (one week). During the simulation, the states (positions, task states) and resource consumption of all agents were recorded at each time step. After the simulation, data analysis was performed: First, the task request frequencies of agent models in each scenario on the spatiotemporal grid were summarized, and hourly spatiotemporal distribution heatmaps of multi-scenario demands were generated using the inverse distance weighting method. Second, by analyzing the concurrency of multiple types of task requests within the same spatiotemporal grid, a conflict period matrix was constructed. The rows and columns of this matrix represent different scenarios, and the matrix element values ​​represent the conflicts between scenarios in specific scenarios. The number or frequency of grids with overlapping demand during a fixed time period is used to quantify the intensity of the conflict. Based on the conflict time period matrix and demand heat map, a collaborative analysis algorithm is executed. For scenarios with highly overlapping demand but adjustable priority within a specific time period, the potential benefits of staggered scheduling are calculated, and specific staggered scheduling suggestions are generated. At the same time, areas that are naturally staggered in time but highly overlapping in space are identified, and the facility demand of the identified areas is extracted into a set of facility demand that can be merged and used. Finally, a structured data set containing candidate locations of shareable facilities, a suggested service time period allocation table, and dynamic demand weight distribution is output as the core demand input for subsequent layout optimization.

[0058] The synergy matrix generation unit learns historical demand data for each scenario based on a deep Q-network, analyzes demand fluctuation patterns, identifies spatiotemporal complementarity between scenarios, generates a scenario synergy matrix and off-peak scheduling strategies, identifies time periods and regions where facilities can be shared across multiple scenarios, and outputs a set of identified sharing candidates. This provides a quantitative basis for multi-scenario facility sharing, improves resource utilization efficiency, achieves efficient sharing of facilities across multiple scenarios, and enhances resource utilization and overall system operational efficiency. It extracts historical demand sequence data for each scenario, uses a deep Q-network to extract temporal features and recognize patterns from the historical demand sequences, analyzes demand fluctuation cycles, intensity changes, and spatial clustering patterns, quantifies spatiotemporal complementarity between scenarios, and through temporal feature extraction, can accurately... Identifying the periodic fluctuation patterns of demand in various scenarios provides a reliable foundation for subsequent complementarity analysis. Based on the spatiotemporal complementarity analysis results, a multi-scenario synergy matrix is ​​constructed. The matrix elements represent the degree of shareability between any two scenarios in terms of time period and region. Based on this, a staggered scheduling strategy and time period allocation scheme are generated. The constructed synergy matrix can intuitively reflect the spatiotemporal sharing potential between scenarios and generate an operational staggered scheduling scheme to improve facility synergy efficiency. Based on the synergy matrix and staggered scheduling strategy, candidate facility areas and time period combinations that meet the conditions for multi-scenario sharing are identified, and a structured shared candidate set is output. The output shared candidate set provides clear candidate area and time period combinations for subsequent site selection and facility configuration, effectively supporting multi-scenario collaborative layout decisions.

[0059] The specific work involves: after analyzing historical spatiotemporal data, extracting features from historical demand sequence data for each scenario, and modeling using a Deep Q-Network (DQN) architecture with the following parameter configurations: the input layer accepts demand intensity sequences formatted as hourly time steps, with a sequence length set to 720 consecutive hours (30 days) to capture complete daily and weekly cycle patterns; the network contains two 256-unit GRU hidden layers for extracting temporal dependency features; the output layer generates 128-dimensional feature vectors through a fully connected network, using a mean squared error loss function and an Adam optimizer (learning rate set to 0.001) during training, with a batch size of 64 and a training cycle of 200 rounds. Then, by analyzing the similarity measure of feature vectors (using cosine similarity calculation), the spatiotemporal complementary patterns between scenarios are accurately identified. Through quantitative analysis, a complementarity coefficient matrix is ​​generated for any two scenarios within a 24-hour cycle on each 100-meter grid unit. Based on the complementarity analysis results, a multi-scenario synergy matrix is ​​constructed. This matrix is ​​an N×N symmetric matrix (N=4, representing four types of scenarios), with matrix elements... This represents the degree of shareability between scene i and scene j within a specific spatiotemporal unit. The time unit is an hour, and the spatial unit is a 100-meter grid. The degree of shareability is calculated unit by unit. At 0.7, the spatiotemporal unit is determined to have high collaborative potential, and a staggered scheduling scheme is automatically generated: For units with time conflicts, the optimal time period offset (offset range ±2 hours) is calculated based on task priority (emergency rescue > government inspection > logistics distribution > tourism) and elasticity coefficient; for units with spatial conflicts, spatial merging is performed based on the facility sharing compatibility matrix (predefined combinations of task types supported by various facilities), and the final output is a complete collaboration degree matrix containing 8760 time units (365 days × 24 hours) and the corresponding staggered scheduling scheme table; the calculation expression of the multi-scenario collaboration degree matrix is ​​as follows:

[0060] ;

[0061] ;

[0062] ;

[0063] In the formula: For the scene With scene The degree of collaboration in a specific spatiotemporal unit; the higher the value, the higher the degree of shareability. The weight of the time complementarity coefficient is 0.6; The weight of the spatial complementarity coefficient is 0.4. This is the time complementarity coefficient. The higher the value, the more complementary the time demand patterns of the two (peaks are staggered). For the scene On a certain 100-meter grid cell, the 128-dimensional feature vector extracted by the DQN-GRU model represents its time demand pattern. For the scene 128-dimensional feature vectors on the same grid cell; This is the spatial complementarity coefficient. The higher the value, the more complementary the two are in terms of spatial demand distribution (hotspot areas are staggered). For the scene Spatial distribution of demand hotspots in the target area within a certain time period (e.g., 1 hour); For the scene Spatial distribution of demand hotspots within the same time period; This represents the area of ​​overlap in the spatial distribution of the two. Scenes The total area of ​​demand hotspots; based on the coordination degree matrix and peak scheduling strategy, the facility sharing condition matching algorithm is executed, and the sharing admission threshold is set: the time dimension requires continuous shareable time period ≥ 4 hours / day, and the spatial dimension requires the coordination grid coverage rate ≥ 30%. All 100-meter grid units are traversed, and the spatiotemporal units that meet the conditions are clustered (using the DBSCAN algorithm) to form candidate sharing areas. Each candidate sharing area generates detailed attributes: spatial range (coordinates of the smallest bounding rectangle), core time period distribution (usage frequency counted by hour), and facility configuration requirements (determining the take-off and landing platform level, number of charging facilities, etc. according to the compatible task type). Finally, a structured sharing candidate set in GeoJSON format is output.

[0064] The constraint fusion assessment module integrates physical information neural networks and multidimensional constraint analysis technology. Based on a shared candidate set, it predicts wind field and noise propagation in complex urban environments. Combining airspace and land use conditions, it constructs an intensity distribution map containing multidimensional constraints, providing high-precision constraint assessment for site selection. The constraint fusion assessment module includes a wind noise environment prediction unit and a multidimensional constraint overlay unit.

[0065] The wind noise environment prediction unit, based on a physical information neural network that integrates fluid dynamics equations and meteorological data, combined with time-period and regional information of shareable facilities in the shared candidate set, simulates urban wind field distribution and noise propagation paths. It predicts wind field distribution and noise propagation under urban terrain, enabling dynamic analysis of wind and noise environmental risks, improving site selection safety and environmental adaptability, accurately identifying wind noise risk areas, and significantly enhancing site selection safety and environmental compliance. It constructs an urban wind field prediction model based on a physical information neural network, integrating the Navier-Stokes equations and high-resolution meteorological data to simulate wind noise in complex terrain. The spatiotemporal distribution of wind speed, wind direction, and turbulence intensity improves the accuracy of low-altitude wind field prediction, effectively supporting the safe site selection of take-off and landing points. Combining aircraft noise source models and acoustic propagation equations, a noise impact range prediction submodule is established. Based on take-off and landing frequency and aircraft parameters, the noise contour distribution is dynamically calculated to achieve a quantitative assessment of noise impact, assisting in the formulation of noise reduction and avoidance strategies. The wind field simulation results and noise propagation predictions are spatiotemporally superimposed to generate a joint risk map, identifying high-risk areas of wind shear and overlapping areas of noise sensitivity, realizing a two-dimensional dynamic assessment of environmental risks, identifying high comprehensive risk areas, and providing environmental constraints for layout optimization.

[0066] The specific work involves constructing a wind field prediction model based on a physical information neural network to accurately simulate wind fields under complex urban terrain. This model uses a feedforward neural network as its core architecture, comprising eight hidden layers, each with 128 neurons, employing the tanh activation function. The input layer receives high-resolution meteorological data and terrain parameters, including: wind speed and direction at 10-meter elevation, temperature gradient, pressure field, three-dimensional vector data of buildings, and surface roughness distribution. The data spatial resolution is 30 meters, and the time step is 15 minutes. During model training, the continuity equation of the Navier-Stokes equations and... The momentum equation is embedded as a physical constraint in the loss function. Network parameters are optimized using a combination of data-driven and physical principle-based approaches. Training data comes from meteorological station data and lidar wind measurement data for the target area over the past three years, employing approximately 100,000 samples. After training, the model can predict the spatiotemporal distribution of wind speed, wind direction, and turbulence intensity at any location within the next 6 hours within a ground-to-150-meter low-altitude range with a root mean square error of less than 0.5 m / s, providing a refined wind environment assessment for takeoff and landing facility site selection. A noise impact prediction model based on acoustic propagation theory is also established, using aircraft... Based on a noise source database, including the sound power level spectrum characteristics of five common aircraft types such as electric vertical takeoff and landing (EVTOL) aircraft and multi-rotor UAVs, the input parameters include engine thrust settings during takeoff and landing, flight trajectory elevation angle, takeoff and landing frequency (flights / hour), and meteorological correction factors. The sound attenuation calculation method according to the ISO 9613-2 standard is adopted, comprehensively considering factors such as geometric divergence, atmospheric absorption, ground effect, and obstacle shielding. The noise impact prediction model establishes a calculation area with a radius of 2 kilometers centered on the candidate takeoff and landing points, with a spatial grid accuracy set to 50 meters × 50 meters. For each calculation point, based on the sound source... The equivalent continuous A-weighted sound level is dynamically calculated based on altitude, propagation distance, and environmental parameters. The model output is a noise contour map of different time periods (peak / off-peak), with the boundaries of the three impact ranges of 55 dB, 65 dB, and 75 dB marked to accurately identify the exceedance of noise standards in noise-sensitive areas such as schools, hospitals, and residential areas. The wind field simulation results are spatiotemporally registered and overlaid with noise prediction data to construct a joint risk classification assessment system. The wind field output is post-processed to extract strong wind areas with wind speeds greater than 8 m / s, turbulent areas with wind direction change rates greater than 30 degrees / 100 m, and vertical wind shear intensity exceeding 0.1 seconds. -1High-risk areas were identified, and areas with sound levels exceeding 65 dB in noise prediction were defined as noise-sensitive areas. Using GIS spatial overlay technology, with a 100m × 100m grid as the evaluation unit, the two types of constraints were weighted and fused: high-risk wind shear areas were assigned a weight of 0.6, and noise-sensitive areas were assigned a weight of 0.4. The contribution of each factor was determined using the analytic hierarchy process (AHP), and a comprehensive risk index was calculated. Combined with preset risk thresholds, a comprehensive risk level distribution map was generated, divided into four levels: low, medium, high, and extremely high. Areas with overlapping high wind and noise risks were automatically marked as prohibited construction zones. For single-risk areas, corresponding protection distance requirements or operational restriction suggestions were proposed based on the risk level, forming a joint risk map of the spatial distribution of constraint intensity.

[0067] The formula for calculating the comprehensive risk index is as follows:

[0068] ;

[0069] ;

[0070] ;

[0071] In the formula: This is a comprehensive risk index; The wind shear risk weight is set to 0.6. This is the noise risk weight, with a value of 0.4; Wind risk intensity index; For the first The weights of each wind field factor are determined using the analytic hierarchy process (AHP). For the first Normalized risk values ​​of individual wind field factors; Noise risk intensity index; , The noise level and sensitivity weights are determined using AHP; The level of noise exceeding the standard, normalized value, is the degree to which it exceeds 65dB; The proximity to sensitive points is a normalized value based on the distance between the grid cell and sensitive points such as schools and hospitals; the risk threshold for low-risk levels is... The risk threshold for medium-risk levels is... The risk threshold for high-risk levels is: The risk threshold for extremely high risk level is: ;

[0072] The multi-dimensional constraint overlay unit integrates multi-dimensional constraints including wind, noise, airspace, and land use. It generates a multi-dimensional constraint intensity distribution map through GIS spatial overlay, forming a visualized constraint map to reduce implementation risks and achieve visualized analysis of constraint conditions. This effectively reduces the risk of planning scheme implementation, extracts and identifies high-risk wind shear areas and noise-sensitive overlapping areas, and integrates airspace control layers, land use data, and building load information to construct a standardized multi-dimensional constraint dataset. This achieves unified storage of multi-source constraint data, providing structurally consistent input for subsequent weighted fusion. Using GIS spatial overlay analysis methods, it performs weighted fusion and conflict detection on various constraint layers, generating a multi-dimensional constraint intensity distribution map that intuitively reflects the suitability and restriction levels of different areas. The constraint intensity distribution map is spatially matched with a shared candidate set to eliminate areas exceeding constraint limits, outputting a suitability score matrix to select feasible candidate areas that meet safety and environmental requirements, providing a quantitative basis for site selection optimization.

[0073] The specific work involves: based on wind field prediction and noise assessment results, extracting high-risk wind shear areas and noise-sensitive overlapping areas to form a preliminary environmental constraint raster layer with a spatial resolution of 100m × 100m; further integrating three types of key constraint data: the airspace control layer, derived from vector data of controlled airspace, airspace protection areas, and temporary restricted areas published by the Civil Aviation Administration, uniformly converted to the WGS84 coordinate system; land use data, based on the city's overall land space plan, extracting ecological protection red lines, urban development boundaries, and vector boundaries of various types of construction land; and building load-bearing information, through integrating existing building structure survey data and roof load-bearing capacity classification maps derived from remote sensing, assigning load-bearing capacity thresholds of no less than 200 kg / m² for light UAV take-off and landing facilities and no less than 500 kg / m² for electric vertical take-off and landing aircraft facilities. All data undergoes spatial registration and rasterization processing to form a multi-dimensional constraint dataset with unified geographic reference and consistent attribute structure. The segment includes constraint type, data source, and update time. A GIS spatial overlay analysis method is used to perform weighted fusion processing on the standardized multidimensional constraint dataset. The weight coefficients for each constraint category are set as follows: environmental constraints (wind shear and noise) weight 0.35, airspace control weight 0.3, land use weight 0.25, and building load-bearing capacity weight 0.1. A weighted overlay analysis algorithm is used to calculate the constraint intensity index grid-by-grid, with a value range of 0-100. Higher values ​​indicate stronger constraints. Conflict detection logic is also executed; when two or more prohibitive constraints exist in the same grid cell, it is automatically marked as a conflicting cell. The final generated multidimensional constraint intensity distribution map contains four constraint level zones: low constraint zone (index 0-30), medium constraint zone (31-60), high constraint zone (61-80), and prohibited construction zone (81-100). The output format is GeoTIFF raster data, and the spatial reference system uses the CGCS2000 Gauss-Krüger projection. For each raster cell… The constraint strength index is calculated using the following expression:

[0074] ;

[0075] ;

[0076] In the formula: The constraint strength index; For the first Weight coefficients for class constraints; For the first Class constraints in grid Normalized constraint values ​​on; These are the original constraint values; and These are the minimum and maximum values ​​for this type of constraint; if the same grid contains ≥2 types of prohibitive constraints, then... Automatically marked as prohibited construction zones; spatial matching analysis is performed between the multidimensional constraint intensity distribution map and the shared candidate set. The shared candidate set is input in GeoJSON format and includes attributes such as candidate area boundaries, suggested service periods, and facility configuration requirements. The constraint intensity values ​​within the coverage area of ​​each candidate area are extracted through spatial join operations. The overall constraint intensity of the area is calculated using an average algorithm. A constraint exceedance threshold of 60 is set, and candidate areas with overall constraint intensity exceeding this threshold are eliminated. The remaining candidate areas are further calculated for suitability scores, and a structured suitability score matrix is ​​finally output. The matrix rows correspond to candidate area numbers, and the columns include spatial center coordinates, constraint intensity values, suitability scores, main limiting factors, and recommended optimization measures. The data is stored in CSV format and linked to the geospatial database. The formula for calculating the suitability score is as follows:

[0077] ;

[0078] ;

[0079] In the formula: This is a suitability score; a higher score indicates greater suitability. Candidate region; For the region The average constraint strength; For the region The number of grid cells covered;

[0080] The multi-objective optimization modeling module combines the scenario synergy matrix, shared candidate set, and intensity distribution map to construct a multi-objective optimization model with the objectives of maximizing service coverage, minimizing system cost, and minimizing negative externalities. It incorporates scenario synergy and constraint data to optimize facility layout under multi-scenario sharing, improving overall system efficiency and fairness. Service coverage is measured by demand fulfillment rate and average response time; system cost includes total lifecycle economic cost; and negative externalities consider noise impact and security risks.

[0081] The dynamic iterative solution module uses the particle swarm optimization algorithm to dynamically iterate and optimize the multi-objective optimization model. It combines real-time constraint updates and demand fluctuations to adjust the layout scheme. It provides a non-dominated solution set through Pareto front analysis, supports multi-scheme comparison and dynamic adjustment, enhances the system's adaptability and foresight, and outputs the optimal optimization scheme that takes into account safety, efficiency and benefits, thereby improving the system's adaptability, foresight and decision-making scientificity.

[0082] The scheme simulation and verification module conducts simulations of optimization schemes in the target area, evaluates the facility operation efficiency, safety and economy under different scenarios, and proposes targeted suggestions and implementation paths from the dimensions of planning, land use and operation based on the simulation results, forming an operable scheme toolkit, verifying the feasibility of the scheme and outputting implementation strategies, thereby improving the operability and success rate of the scheme.

[0083] Example 2, as Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: The execution steps of the multi-objective optimization modeling module are as follows: With the objectives of maximizing service coverage, minimizing system cost, and minimizing negative externalities, a multi-objective optimization model is constructed. Service coverage is measured by demand satisfaction rate and average response time. System cost includes the full life cycle economic cost. Negative externalities consider noise impact and safety risks. A multi-objective optimization model containing three major objectives is constructed to achieve a three-dimensional balanced decision on service efficiency, economic benefits, and safety impact. The scenario coordination degree matrix, shared candidate set, and multi-dimensional constraint strength distribution map are used as model inputs. The decision variables are defined as facility location, type, service period, and shared allocation scheme. Combining coordination degree, candidate set, and constraint strength map, the decision variables are quantified to provide structured input for dynamic layout optimization. Model constraints are set, including demand coverage threshold, maximum construction cost, noise safety limit, and airspace availability, so that the scheme meets actual operation and safety requirements. Through the embedding of hard constraints, the scheme is ensured to have practical operability, taking into account both safety compliance and cost control.

[0084] The specific work involves: constructing a multi-objective optimization model with the objectives of maximizing service coverage, minimizing system cost, and minimizing negative externalities. The service coverage objective uses demand satisfaction rate (≥85%) and average response time (≤15 minutes) as quantitative indicators. The system cost objective covers the entire lifecycle economic cost during the construction and operation phases. Construction costs are estimated by facility type (simple type: 500,000-800,000 RMB; standard type: 1,500,000-3,000,000 RMB; comprehensive type: 5,000,000-10,000,000 RMB). Operation costs are calculated annually based on service scale (approximately 5-8% of construction costs). The negative externalities objective quantifies noise impact (equivalent continuous A-weighted sound level limit: daytime ≤65dB, nighttime ≤55dB) and safety risks (wind shear intensity threshold ≤0.1s). -1The model input includes three types of data: a scenario coordination degree matrix, a shared candidate set, and a multi-dimensional constraint strength distribution map. All input data undergoes spatial registration and standardization to ensure coordinate system consistency and data format compatibility. The model defines four types of decision variables: facility location (a binary variable representing whether to construct a take-off and landing point in the candidate area); facility type (a discrete variable with values ​​of 1, 2, and 3, corresponding to simple, standard, and comprehensive facilities, respectively); service time (a continuous variable divided by hours, representing the scenario types served by the facility at different times); and a shared allocation scheme (a multi-dimensional matrix describing the scenario combinations and resource allocation ratios of each facility within a specific time period). Constraints include four types of hard restrictions: demand coverage constraint requiring a demand satisfaction rate of no less than 70% for each 100-meter grid unit; maximum construction cost constraint setting an upper limit based on the investment budget (e.g., total investment in a single area not exceeding 50 million yuan); and noise safety limit constraint requiring facility operating noise at sensitive points not to exceed a specified threshold (learning...). Strict limits are enforced in areas such as schools and hospitals; airspace availability constraints require that facility locations not overlap with controlled airspace or protected areas, and must meet minimum safety interval requirements (horizontal interval ≥ 500 meters, vertical interval ≥ 50 meters). All constraints are embedded in the model through linear or nonlinear inequalities. A multi-objective particle swarm optimization algorithm is used to solve the model, with a population size of 200, a maximum number of iterations of 500, and an inertia weight dynamically adjusted from 0.4 to 0.9. During the solution process, the scenario synergy matrix and constraint intensity distribution map are read in real time, and the facility layout and sharing scheme are dynamically adjusted. Pareto front analysis is used to generate a non-dominated solution set. The output schemes include: facility site layout map (vector format), time-segmented service allocation table, full life cycle cost budget table, and negative externality impact assessment report. All outputs are managed in association through a spatial database, supporting scheme comparison and parameter sensitivity analysis. The final scheme meets all constraints and achieves an equilibrium state among service coverage, cost control, and risk mitigation.

[0085] The execution steps of the dynamic iterative solution module are as follows: The multi-objective optimization model is solved using the particle swarm optimization algorithm, with embedded constraint handling mechanisms and dynamic inertia weighting strategies. This supports efficient optimization in complex solution spaces, significantly improving the algorithm's convergence speed and stably approaching the Pareto front within 500 generations. During the iteration process, updated demand data and constraint information are integrated in real time, dynamically adjusting facility layout schemes. Non-dominated solution sets are selected through Pareto sorting, enabling layout schemes to dynamically adapt to the environment and ensuring the real-time feasibility and robustness of the solution set. Diversity preservation and convergence analysis are performed on the Pareto front solution set, outputting several optimal optimization schemes that balance safety, efficiency, and benefits. Multiple sets of equilibrium optimization schemes are provided, allowing decision-makers to flexibly choose implementation paths based on actual preferences.

[0086] The specific work involves: using particle swarm optimization (PSO) to construct a dynamic search mechanism with constraint handling capabilities to solve a multi-objective optimization model. During algorithm initialization, the population size is set to 200, and the particle dimension is determined by the total number of decision variables, including dimensions such as facility location, type, and time period allocation. Each particle's position vector corresponds to a complete layout scheme, and the velocity vector is initialized to a random value within the range [-1, 1]. A penalty function method is then used to transform the degree of constraint violation into a penalty term in the objective function, with the penalty coefficient for hard constraints set to 10. 6 The penalty coefficient for soft constraints is set to 10. 3 The dynamic inertia weight adopts a linear decreasing strategy, gradually decreasing from 0.9 to 0.4. The learning factors c1 and c2 are both set to 2.0, and the maximum number of iterations is set to 500 generations. In each iteration, the fitness value of each particle is calculated through the evaluation function, which includes three normalized objective function values ​​and constraint penalty terms, ensuring that the search process always takes place near the feasible region. During the optimization iteration process, a real-time data access interface is established, and updated demand data and constraint information are read every 50 generations. The demand data is updated with a 15-minute granularity and includes the dynamic demand distribution of four scenarios. The constraint information is updated every 6 hours, covering wind field prediction results and changes in airspace status. When a data update is detected, the constraint matrix and objective function value of the affected area are recalculated, and the particle velocity is adjusted through an adaptive mechanism: if the constraints tighten, the velocity of the corresponding dimension is reduced by 20%-40%; if the demand hotspot shifts, the particle accelerates to the new hotspot area by 0.1-0.3 units of velocity. After each iteration, the population is Pareto stratification is performed using the fast non-dominated sorting method, and the top 10 are selected. Zero non-dominated solutions are stored in the elite archive. The archive maintenance uses a crowding comparison operator to ensure the uniform distribution of the solution set in the target space. The crowding threshold is set to 0.01. At the same time, the epsilon dominance strategy is implemented, with the epsilon value set to 0.001 to eliminate approximately duplicate solutions in the target space. After the algorithm converges, a systematic analysis is performed on the final Pareto front solution set. The spacing index of the solutions is calculated to evaluate the uniformity of distribution, requiring the spacing value to be less than 0.1. The degree of convergence is measured by the hypervolume index. The hypervolume index is analyzed using an ideal reference point (1.1, 1.1, 1.1). The solutions with the highest contribution of the hypervolume index are retained. At the same time, the K-means clustering method is used to divide the solution set into 3-5 groups. The clustering features include the number of facilities, average service radius, and total construction cost. The solution closest to the cluster center is selected from each cluster as the representative output. Each output solution includes a complete list of facility coordinates, a time-segmented scheduling table, a detailed cost composition, and a risk quantification assessment. All data are stored in the spatial database in GeoJSON and CSV formats.

[0087] The execution steps of the scheme simulation and verification module are as follows: Based on the output optimal optimization scheme, a discrete event simulation model is built in the target area to simulate the dynamic scheduling and queuing process of multiple scenario tasks in the facility network, so as to realize the refined reproduction of the scheme operation process, expose potential conflicts and bottlenecks in advance, set up multiple operation scenarios, and simulate and evaluate verification indicators including facility utilization rate, average response time, noise impact range, wind shear accident probability and full-cycle economic cost. The robustness and economic feasibility of the scheme under various future conditions are quantitatively evaluated. Based on the simulation results, system bottlenecks and potential risks are identified. Implementation suggestions and dynamic adjustment strategies are generated from three dimensions: planning connection, land use coordination and operation management. A deliverable optimization scheme toolkit is formed, providing implementable adjustment measures and control rules to improve the actual operability of the scheme.

[0088] The specific work involves: constructing a discrete event simulation model of the target area based on the facility coordinate list and service allocation table output by the optimization scheme. The model uses AnyLogic 9.8 as the simulation platform and defines four types of task generators: emergency rescue tasks follow a Poisson distribution (λ=0.5 times / hour), logistics and distribution tasks are generated based on historical order time-series clustering (1.2 times / hour during peak hours), tourism and sightseeing tasks are dynamically adjusted based on scenic spot visitor flow data, and government inspection tasks are periodically triggered according to a fixed inspection plan (4 times / day). Limited-capacity queues are set up at each take-off and landing point in the facility network, using priority scheduling rules (emergency tasks first, logistics second), and physical resource constraints are configured: the maximum simultaneous service for simple facilities is 2 flights, for standard facilities it is 4 flights, and for comprehensive facilities it is 8 flights. The simulation time step is set to 1 minute. The simulation period is 30 consecutive days (including 4 full weekends). During initialization, a geospatial base map and facility topology are loaded to ensure a unified spatiotemporal reference. Three typical operational scenarios are set up for simulation: a baseline scenario (demand for each scenario is based on historical averages), a growth scenario (logistics and tourism demand increase by 15% annually), and an extreme scenario (emergency event frequency increases by 50%). During the simulation, five types of verification indicators are collected in real time: facility utilization rate is statistically analyzed at 15-minute intervals (threshold >75% is considered overload); average response time is calculated as the interval from task issuance to landing (emergency ≤15 minutes, logistics ≤25 minutes); noise impact range is dynamically output using an integrated acoustic model with 55 / 65 / 75 dB isopleths; and wind shear accident probability is based on real-time wind field data and flight envelope thresholds (wind speed >12 m / s or shear >0.1 s). -1The simulation process involves triggering risk counts, accumulating the full-cycle economic cost at an annual discount rate of 5% for construction and operation expenditures, storing the simulation results in a structured database, and supporting multi-dimensional comparative analysis and visual dashboard displays. Based on the simulation output's time-series data and spatial distribution map, system bottlenecks are identified: excessive queue length for logistics and tourism tasks in core business districts during peak hours (>6 flights), excessive probability of strong wind and daytime wind shear risk (>0.05 times / day), and utilization rates of some simple facilities consistently below 30%. Based on this, implementation strategies are generated from three dimensions: in terms of planning coordination, it is recommended to add backup take-off and landing points within a 1-kilometer radius of areas with utilization rates >80% (land use compatible with commercial / transportation); in terms of land use coordination, time-limited restrictions are proposed for facilities around noise-sensitive areas (take-off and landing are prohibited from 22:00 to 6:00 at night); in terms of operation management, a dynamic scheduling rule base is established, automatically switching to wind-resistant backup landing points when wind speed forecast >10m / s. Finally, a solution toolkit containing simulation reports, adjustment strategy lists, and control rule configuration files is formed.

[0089] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0090] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A spatiotemporal constraint analysis system for the collaborative layout of urban low-altitude multi-scenario take-off and landing facilities, comprising a cloud management center, characterized in that, The cloud management center communication connection includes the following modules: The scene collaboration perception module comprehensively utilizes multi-agent reinforcement learning and deep Q-network to simulate the spatiotemporal demand dynamics of four types of scenes in the target area, generate a scene collaboration degree matrix and a staggered scheduling strategy, identify the time period and area of ​​shareable facilities, and output a sharing candidate set. The constraint fusion assessment module integrates physical information neural networks and multidimensional constraint analysis technology. Based on a shared candidate set, it predicts wind field and noise propagation in complex urban environments and constructs an intensity distribution map containing multidimensional constraints by combining airspace and land use conditions. The multi-objective optimization modeling module is used to combine the scenario synergy matrix, shared candidate set and intensity distribution map to build a multi-objective optimization model with the objectives of maximizing service coverage, minimizing system cost and minimizing negative externalities, and incorporates scenario synergy and constraint data. The dynamic iterative solution module uses the particle swarm optimization algorithm to dynamically iteratively optimize the multi-objective optimization model. It combines real-time constraint updates and demand fluctuations to adjust the layout scheme. It provides a non-dominated solution set through Pareto front analysis and outputs the optimal optimization scheme that takes into account safety, efficiency and benefits. The scheme simulation and verification module conducts simulations of optimization schemes in the target area, and proposes targeted suggestions and implementation paths from the dimensions of planning, land use and operation based on the simulation results, forming an operable scheme toolkit.

2. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 1, characterized in that: The scene collaborative perception module includes an intelligent agent interaction simulation unit and a collaborative degree matrix generation unit; The intelligent agent interaction simulation unit is used to introduce a multi-agent reinforcement learning framework into the target area, construct an intelligent agent model covering four scenarios: emergency rescue, logistics distribution, tourism and government inspection, simulate the demand interaction and fluctuation patterns in the spatiotemporal dimension, and identify demand hotspots and conflict periods. The synergy matrix generation unit learns historical demand data for each scenario based on a deep Q-network, analyzes demand fluctuation patterns and identifies spatiotemporal complementarity between scenarios, generates a scenario synergy matrix and a staggered scheduling strategy, identifies time periods and regions where facilities can be shared among multiple scenarios, and outputs the identified shared candidate set.

3. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 2, characterized in that: The execution steps of the intelligent agent interaction simulation unit are as follows: Based on the city's overall plan or pilot areas for low-altitude economic development, target areas are identified, and historical spatiotemporal data covering four scenarios—emergency rescue, logistics and distribution, tourism and sightseeing, and government inspection—are obtained. Construct intelligent agent models for four scenarios in the target area. Each intelligent agent model models demand behavior based on historical spatiotemporal data, simulating its task triggering logic and movement patterns within the target area. A multi-agent reinforcement learning framework is introduced, and interaction rules and resource competition mechanisms between agent models are set. Through time-series simulation, demand conflicts and overlaps in different scenarios during peak hours and hot spots are identified. Based on simulation results, a heat map of the spatiotemporal distribution of scenario requirements and a matrix of conflict periods are dynamically generated, and a set of facility requirements that can be merged or used in staggered periods is extracted.

4. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 2, characterized in that: The execution steps of the coordination degree matrix generation unit are as follows: Historical demand sequence data of each scenario are extracted, and a deep Q-network is used to extract temporal features and recognize patterns of historical demand sequences of each scenario. The cycle, intensity and spatial clustering of demand fluctuations are analyzed, and the spatiotemporal complementarity between scenarios is quantified. Based on the spatiotemporal complementarity analysis results, a multi-scenario synergy matrix is ​​constructed. The matrix elements represent the degree of shareability between any two scenarios in terms of time period and region, and a staggered scheduling strategy and time period allocation scheme are generated accordingly. Based on the synergy matrix and the off-peak scheduling strategy, candidate facility areas and time slot combinations that meet the conditions for sharing in multiple scenarios are identified, and a structured shared candidate set is output.

5. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 2, characterized in that: The constraint fusion evaluation module includes a wind noise environment prediction unit and a multi-dimensional constraint overlay unit; The wind noise environment prediction unit, based on the physics information neural network, integrates fluid dynamics equations and meteorological data, and combines time period and regional information of shareable facilities in the shared candidate set to simulate urban wind field distribution and noise propagation path, and predict wind field distribution and noise propagation under urban terrain. The multidimensional constraint overlay unit is used to integrate multidimensional constraints including wind, noise, airspace, and land use, and generate an intensity distribution map of multidimensional constraints through GIS spatial overlay to form a visualized constraint map.

6. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 5, characterized in that: The execution steps of the wind noise environment prediction unit are as follows: A city wind field prediction model is constructed based on physical information neural network, which integrates Navier-Stokes equation and high-resolution meteorological data to simulate the spatiotemporal distribution of wind speed, wind direction and turbulence intensity under complex terrain. By combining the aircraft noise source model and the acoustic propagation equation, a noise impact range prediction submodule is established, and the noise contour distribution is dynamically calculated based on the takeoff and landing frequency and aircraft parameters. By spatiotemporally overlaying wind field simulation results with noise propagation predictions, a joint risk map is generated to identify high-risk areas of wind shear and overlapping areas of noise sensitivity.

7. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 5, characterized in that: The execution steps of the multidimensional constraint overlay unit are as follows: Extract the identified high-risk wind shear areas and noise-sensitive overlapping areas, and integrate airspace control layers, land use data, and building load information to construct a standardized multidimensional constraint dataset; Using GIS spatial overlay analysis, weighted fusion and conflict detection are performed on various constraint layers to generate a multidimensional constraint intensity distribution map. Spatial matching is performed between the constraint intensity distribution map and the shared candidate set to eliminate regions where constraints exceed the limits, and a suitability score matrix is ​​output.

8. The spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 5, characterized in that: The execution steps of the multi-objective optimization modeling module are as follows: With the goals of maximizing service coverage, minimizing system cost, and minimizing negative externalities, a multi-objective optimization model is constructed. Service coverage is measured by demand satisfaction rate and average response time, system cost includes the total lifecycle economic cost, and negative externalities consider noise impact and security risks. The scenario synergy matrix, shared candidate set, and multidimensional constraint strength distribution map are used as model inputs, and the decision variables are defined as facility location, type, service period, and shared allocation scheme. Set model constraints, including demand coverage threshold, maximum construction cost, noise safety limit, and airspace availability, to ensure that the solution meets actual operation and safety requirements.

9. A spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 8, characterized in that: The execution steps of the dynamic iterative solution module are as follows: The particle swarm optimization algorithm is used to solve the multi-objective optimization model, and a constraint handling mechanism and dynamic inertia weight strategy are embedded. During the iteration process, the updated demand data and constraint information are accessed in real time, the facility layout scheme is dynamically adjusted, and the non-dominated solution set is selected by Pareto sorting. A diversity preservation and convergence analysis is performed on the Pareto front solution set to output several optimal solutions that balance safety, efficiency, and effectiveness.

10. A spatiotemporal constraint analysis system for collaborative layout of urban low-altitude multi-scenario take-off and landing facilities according to claim 9, characterized in that: The execution steps of the scheme deduction and verification module are as follows: Based on the optimal optimization scheme of the output, a discrete event simulation model is constructed in the target area to simulate the dynamic scheduling and queuing process of multiple scenario tasks in the facility network. Multiple operational scenarios were set up to simulate and evaluate verification indicators including facility utilization, average response time, noise impact range, wind shear accident probability, and full-cycle economic cost. Based on the simulation results, system bottlenecks and potential risks are identified, and implementation suggestions and dynamic adjustment strategies are generated from three dimensions: planning integration, land use coordination, and operation management, forming a deliverable optimization solution toolkit.

Images