An eVTOL urban air emergency medical landing point layout method considering demand point clustering

Abstract

This invention belongs to the field of urban aviation technology and discloses an eVTOL urban aviation emergency medical take-off and landing point layout method that considers demand point clustering. This invention identifies and extracts potential demand points for urban medical emergencies, uses the DBSCAN algorithm to perform spatial clustering and dimensionality reduction on these demand points, and defines urban aviation medical emergency rescue operation scenarios as two typical scenarios based on the essential differences in emergency rescue missions. Subsequently, a maximum coverage location optimization model (MCLP) considering multiple constraints is constructed, and an improved genetic algorithm is used to solve the model, ultimately generating a hierarchical take-off and landing point layout scheme and service network. This method can achieve 100% coverage of demand points throughout the city under fixed resources. DBSCAN clustering effectively solves the redundancy problem of the original POI data, improves the accuracy of layout decisions, and fully considers the different operational logics of material distribution and casualty transfer, thus better meeting the needs of real medical rescue.

Description

Technical Field

[0001] This invention relates to the field of urban aviation technology, and more specifically, to an eVTOL urban aviation emergency medical take-off and landing point layout method that takes into account demand point clustering. Background Technology

[0002] In the construction of an urban aviation emergency medical rescue system, the layout of take-off and landing points is the core foundation and key prerequisite. The rationality of its layout directly determines the coverage, response efficiency and service accessibility of the rescue network.

[0003] Currently, urban air medical rescue mainly relies on helicopters, but their operating costs are high, they are noisy, and the number of take-off and landing points is severely insufficient, making it difficult to support high-frequency urban rescue operations. In terms of take-off and landing point layout research, existing studies mostly focus on ordinary helicopters or logistics drones, lacking comprehensive optimization models that consider the performance characteristics of eVTOL (such as endurance, payload, and vertical take-off and landing requirements) and urban medical emergency scenarios (highly time-sensitive, with unevenly distributed resources and space). Furthermore, original points of medical need (POIs) are often redundant and highly unevenly distributed; directly performing model calculations leads to inefficient and impractical solutions. Traditional urban air emergency medical take-off and landing point layout methods do not fully consider the actual distribution characteristics of emergency needs within urban space, and do not conduct refined spatial screening and optimization of potential demand points, resulting in high redundancy and scattered distribution of demand input data. Traditional layout methods directly output a single site selection result without analyzing the spatial distribution characteristics of take-off and landing points, the service coverage effect, or developing differentiated hierarchical layout strategies for areas with different demand intensities. It is impossible to efficiently guarantee the timeliness of emergency medical care in the core urban areas and it is difficult to fill the service gaps in peripheral areas, thus making it difficult to achieve a balanced allocation and efficient operation and maintenance of emergency medical rescue resources throughout the city. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a method for the layout of eVTOL urban aviation emergency medical take-off and landing points that considers demand point clustering.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] S1. Identify and extract potential demand points for urban medical emergencies. Use urban point of interest data as spatial proxy variables for potential medical emergency demands. Select functional types related to the occurrence of medical emergency events as raw data. Clean, deduplicate, and transform the coordinates of the data to construct an initial set of potential demand points for urban medical emergencies.

[0007] S2. Based on the DBSCAN algorithm, spatial clustering and dimensionality reduction are performed on the demand points. A density-based spatial clustering algorithm is used to process the initial potential medical emergency demand points. A distance metric method suitable for geospatial data is introduced to calculate the distance relationship between points. By setting the neighborhood radius and minimum sample number parameters, the initial potential medical emergency demand points in space are clustered, and the discretely distributed initial potential medical emergency demand points are aggregated into several demand clusters. The center point of each demand cluster is extracted as the final rescue demand point. At the same time, noise points are removed, thereby reducing the data scale and computational complexity while preserving the spatial distribution characteristics of the demand.

[0008] The neighborhood radius α is determined as follows: For each initial potential medical emergency demand point, calculate its distance to the k-th nearest neighbor point, sort all these distance values ​​in ascending order, and plot a k-distance curve. The distance value corresponding to the inflection point where the curve changes from a gradual change to a steep increase is taken as the neighborhood radius. ɛ The reference value is determined by taking into account the data scale and spatial distribution characteristics; the minimum sample size parameter is determined by taking into account the data scale and spatial distribution characteristics.

[0009] S3. Define typical urban air medical emergency rescue operation scenarios. Based on the essential differences in service objectives, time constraints, and operation paths of eVTOL urban medical emergency rescue missions, eVTOL urban medical emergency rescue missions are divided into two typical scenarios: direct medical rescue and indirect medical transport. Direct medical rescue is a response-oriented scenario, focusing on the delivery of emergency supplies or the arrival of emergency personnel. The operation mode is hospital take-off / landing point—demand point—return to the original hospital. The constraint is that the round-trip flight time and distance must meet the threshold. The second scenario is a transport-oriented scenario, focusing on transferring patients from the scene to the hospital. The operation mode is hospital take-off / landing point—demand point—tertiary hospital—original hospital. This scenario does not limit the return to the original take-off point.

[0010] S4. Construct a maximum coverage location model (MCLP) that considers multiple constraints. With the goal of maximizing the effective coverage of medical emergency needs, construct a take-off and landing point location model. In the model, comprehensively consider the limitations on the number of take-off and landing points, flight response time constraints, flight range constraints, and service capacity constraints of individual take-off and landing points. Establish the coverage relationship between the final rescue demand point and candidate take-off and landing points, and form a complete mathematical programming model to characterize the relationship between take-off and landing point layout and demand coverage.

[0011] S5. An improved genetic algorithm is used to solve the model. For the constructed site selection model, an improved genetic algorithm is introduced for solution. Coverage relationship preprocessing reduces computational complexity, and constraint embedding and solution repair mechanisms ensure the feasibility of the solution. Under the premise of satisfying the constraints, the take-off and landing point layout scheme is solved. The specific steps are as follows:

[0012] S51. To address the characteristics of the location selection problem, a genetic algorithm (GA) is designed to solve it. The binary encoding is used to represent the location status of the take-off and landing points. The fitness function is designed to cover the total demand. The algorithm iteratively searches through selection, crossover, and mutation operations until the termination condition is met, and then outputs the location selection scheme.

[0013] The fitness function is defined as:

[0014]

[0015] Where F represents the individual fitness value;

[0016] S52. Selection operator: Using a roulette wheel selection method based on fitness ratios, after calculating the fitness of individuals in the population, individuals need to be selected from the current population to enter the next generation.

[0017] ① The fitness of each individual in the population is nonnegated:

[0018]

[0019] in, This represents the initial fitness value of the k-th individual; Indicates the first The nonnegated fitness value of each individual; min(f) represents the minimum fitness value of all individuals in the current population;

[0020] ② Calculate the individual's choice probability:

[0021]

[0022] in, This represents the probability that the k-th individual is selected; N represents the population size.

[0023] ③ Construct the cumulative probability distribution:

[0024]

[0025] Where, q k p represents the cumulative selection probability corresponding to the k-th individual; t This represents the selection probability of the t-th individual; t represents the individual index.

[0026] ④ Generate a random number r within the interval [0,1], and select one that satisfies: Individuals enter the next generation of the population;

[0027] S53. Single-point crossover is adopted, combined with a solution repair mechanism, to ensure that the generated individuals always meet the facility quantity constraint and to ensure that the algorithm search process always takes place within the feasible solution space. The specific steps are as follows:

[0028] ① Randomly select parent individuals based on crossover probability Pc; ② Randomly select crossover point c within the chromosome length range; ③ Exchange gene segments after the crossover point between the two individuals to generate offspring individuals; ④ Perform feasibility repair on the individuals after crossover:

[0029] like Randomly set any extra "1"s to "0"s;

[0030] like Randomly set some "0"s to "1"s until the constraint is satisfied;

[0031] Where j represents the candidate take-off and landing point; x j This indicates whether candidate take-off and landing point j has been selected for construction; P represents the maximum number of take-off and landing points allowed to be constructed during the planning period.

[0032] S6. Analysis method of take-off and landing point layout results based on site selection results: By analyzing the spatial distribution characteristics and overall coverage effect of take-off and landing points, key nodes can be identified and laid out.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] (1) Improve rescue efficiency: Compared with the traditional layout, this method can achieve 100% coverage of the city's demand points with fixed resources.

[0035] (2) Data scientificity: DBSCAN clustering effectively solved the redundancy problem of the original POI data and improved the accuracy of layout decisions.

[0036] (3) Strong scenario adaptability: It fully considers the different operational logics of material distribution and patient transfer, which is more in line with the real medical rescue needs. Attached Figure Description

[0037] Figure 1 This invention provides a schematic diagram illustrating the steps of an eVTOL urban aviation emergency medical take-off and landing point layout method that considers demand point clustering in an embodiment of the present invention.

[0038] Figure 2 This is a schematic diagram illustrating the demand point clustering effect of an eVTOL urban aviation emergency medical take-off and landing point layout method that considers demand point clustering, provided in an embodiment of the present invention.

[0039] Figure 3 This invention provides a layout method for eVTOL urban aviation emergency medical take-off and landing points that considers demand point clustering, and the hospital, demand point, and original hospital round-trip pattern site selection layout results are shown in the figure.

[0040] Figure 4This invention provides a layout result diagram of hospitals, demand points, and the nearest hospital transfer pattern in an eVTOL urban aviation emergency medical take-off and landing point layout method that considers demand point clustering; wherein, Figure 4 In this context, 'a' represents the global service network. Figure 4 In the middle, b represents the matching logic for a single demand point. Detailed Implementation

[0041] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to specific examples. Figure 1 As shown, the specific implementation steps of the present invention are as follows:

[0042] S1. Identify and extract potential medical emergency demand points, anchoring the potential service recipients and demand sources of air emergency medical rescue from the urban spatial perspective, and using urban points of interest (POI) data as spatial proxy variables for potential medical emergency demand. Select functional types highly correlated with the occurrence of medical emergency events, such as residential areas, schools, parks, and commercial centers, as the raw data. Clean, deduplicate, and transform the coordinates of the data to construct an initial set of potential medical emergency demand points.

[0043] Through the Gaode Map Open Platform, POI location data for commercial areas, industrial areas, residential areas, tourist attractions, and school areas within xx city were obtained. After data cleaning, a total of 479 data points were obtained for commercial areas, 220 for industrial areas, 299 for residential areas, 590 for tourist attractions, and 519 for school areas. The corresponding demand points for each demand area are shown in the table below:

[0044] Table 1. Demand Data for Commercial Areas

[0045]

[0046] Table 2 Demand Data for Industrial Zones

[0047]

[0048] Table 3. Residential Area Demand Data

[0049]

[0050] Table 4. Demand Data for Tourist Attractions

[0051]

[0052] Table 5 School District Demand Data

[0053]

[0054] In this study, the candidate set of landing and takeoff points was constructed based on medical institutions in XX City with emergency medical rescue capabilities. After systematically cleaning and organizing the raw data, removing duplicates and invalid information, 113 candidate points were obtained, as detailed in Table 6. Furthermore, considering the characteristics of the Scenario 2 (original hospital—demand point—tertiary hospital—return to original hospital) operation mode, to ensure that the receiving end medical resources have high emergency response capabilities and service levels, 30 tertiary hospitals in XX City were selected from the above candidate set as the landing and takeoff point candidate set, with relevant data detailed in Table 7. This tiered selection process ensured both the spatial coverage of the candidate points and took into account the differentiated needs for emergency medical service capabilities under different operation modes.

[0055] Table 6. Candidate Takeoff and Landing Point Data

[0056]

[0057] Table 7. Candidate Set of Landing Points for Grade III Class A Hospitals in XX City (Selected)

[0058]

[0059] S2. Spatial clustering and dimensionality reduction of initial potential medical emergency demand points based on the DBSCAN algorithm. A density-based spatial clustering algorithm is used to process the initial potential medical emergency demand points. A distance metric suitable for geospatial data is introduced to calculate the distance relationship between points. By setting neighborhood radius and minimum sample size parameters, initial potential medical emergency demand points with high spatial density are clustered, aggregating the discretely distributed initial potential medical emergency demand points into several demand clusters. The center point of each demand cluster is extracted as the final rescue demand point, while noise points are removed. This reduces the data scale and computational complexity while preserving the spatial distribution characteristics of the demand. The specific steps are as follows:

[0060] S21. The Haversine Distance is introduced as a distance metric to eliminate the bias caused by the curvature of the Earth.

[0061] To address the issue of large volume and uneven distribution of the original POI data, a density-based spatial clustering algorithm, DBSCAN, is used for processing. Let the initial set of potential medical emergency demand points be:

[0062]

[0063] Among them, s m This represents the m-th initial potential medical emergency demand point, whose spatial location is represented by two-dimensional coordinates (lon). m lat m )express.

[0064] In the DBSCAN algorithm, the formation of the cluster structure mainly depends on the following two key parameters: neighborhood radius ɛ and minimum number of samples Minpts.

[0065] For any initial potential medical emergency demand point s m Its neighborhood radius ɛ is defined as:

[0066]

[0067] Where, N ɛ (s m ) represents all pairs of s m The set of sample points whose distance does not exceed ɛ; dist(·) represents the spatial distance between the initial potential medical emergency demand points, s q This represents any initial potential demand point for medical emergencies in the dataset.

[0068] Traditional Euclidean distance cannot eliminate projection bias caused by the curvature of the Earth. To ensure the geographic accuracy of spatial clustering, this study uses Haverthan distance as a metric for neighborhood determination. For any two points s1(φ1, λ1) in the dataset, the Haverthan distance is used. The formula for calculating the Haversian distance d is as follows:

[0069]

[0070] In the formula, φ and λ represent the radian values ​​of latitude and longitude, respectively; R is the average radius of the Earth, which is taken as 6371 km in this paper.

[0071] S22. Determine the optimal neighborhood radius ɛ and the minimum number of samples MinPts using the k-distance graph.

[0072] (1) Setting the minimum sample size MinPts: Generally speaking, the value of MinPts is closely related to the data dimension. For two-dimensional spatial point data, existing studies generally recommend MinPts ≥ 4. Considering that the demand points for urban air medical emergency rescue have a certain degree of spatial clustering, and at the same time, it is necessary to avoid misjudging a small number of accidentally adjacent points as effective clusters, based on the comprehensive data scale and spatial distribution characteristics, MinPts is set to a relatively conservative value to ensure that the identified clustered areas have a certain degree of spatial stability and practical significance.

[0073] (2) Setting the neighborhood radius ɛ: The neighborhood radius ɛ is used to characterize the spatial density of the initial potential medical emergency demand points and is a key parameter affecting the clustering results. If the value of ɛ is too small, a large number of demand points will not be able to form an effective cluster; if the value is too large, it may merge the initial potential medical emergency demand points that are originally spatially different into the same cluster.

[0074] To avoid biases caused by subjective parameter settings, a method based on the k-distance curve is used to assist in determining ɛ. Specifically, for each initial potential medical emergency demand point, the distance to its k-th nearest neighbor (where k = MinPts) is calculated, and these distance values ​​are sorted in ascending order to plot a k-distance curve. The distance value corresponding to the inflection point where the curve transitions from a gradual change to a sharp rise can be used as a reference value for the neighborhood radius ɛ.

[0075] S23. Perform clustering to aggregate high-density areas into core demand clusters, extract the cluster centers as the final rescue demand points for model input, and eliminate noise points, thereby significantly reducing the data size while preserving spatial structure features.

[0076] By clustering the data using the DBSCAN clustering algorithm, the k-distance curve shows a clear inflection point around a distance of approximately 2, and the slope of the curve increases significantly after this value, indicating a change in the spatial density structure. Therefore, ɛ≈2 is used as the initial reference value. Considering that too small a value of ɛ will lead to fragmentation of the clustering results and an increase in the number of noise points, while too large a value of ɛ will cause excessive merging between different clusters, this paper performs multiple adjustments and comparative analyses of the parameters based on the initial estimate, and finally determines the DBSCAN clustering parameters as ε=3.456 and MinPts=4.

[0077] Figure 2 As shown, under the above parameter settings, cluster analysis was performed on 2107 initial potential medical emergency demand points, ultimately forming 41 demand point clusters. Each cluster center can well represent the spatial distribution characteristics of the potential emergency medical demand around it. The corresponding coordinates of the initial potential medical emergency demand point cluster centers are shown in Table 8. In the subsequent maximum coverage location model, these 41 cluster centers were used as the final rescue demand points for modeling and optimization calculations.

[0078] Table 8 Results of Clustering of Initial Potential Medical Emergency Needs

[0079]

[0080] S3. Define typical urban air medical emergency rescue operation scenarios. In the urban medical emergency rescue system, eVTOL, with its strong vertical takeoff and landing capabilities, flexible deployment, fast response speed, and low dependence on ground transportation, can quickly reach the scene of an incident to provide emergency medical care, and can also undertake cross-regional medical transfer missions, thus possessing wide applicability in various emergency scenarios. However, different mission types have fundamental differences in service objectives, time constraints, and operational paths. Specifically, direct medical rescue missions oriented towards on-site treatment rely on eVTOL's rapid response and point-to-point arrival capabilities, emphasizing the "rapid arrival—on-site treatment—return from the origin" operational mode, and are highly sensitive to response time. Indirect medical transfer missions, centered on the transfer of patients or critical medical resources, rely more on eVTOL's ability to achieve efficient cross-regional transportation under complex traffic conditions. Their operation involves spatial transfer from "demand point—medical institution," placing higher demands on route organization and medical resource matching. Due to these differences, the two types of missions have significant differences in flight path structure, service processes, and constraints. Based on this, this paper divides eVTOL urban medical emergency rescue missions into two typical scenarios: direct medical rescue and indirect medical transport, to support subsequent differentiated modeling and optimization analysis, thereby more accurately evaluating the system's response efficiency and service capabilities in different application scenarios.

[0081] Based on the nature of the mission, two typical scenarios are categorized: Scenario 1 is rapid response type, focusing on the delivery of emergency supplies or the rapid arrival of emergency personnel. The operational mode is "hospital take-off / landing point—demand point—return to the original hospital," with constraints that the round-trip flight time and distance must meet thresholds; Scenario 2 is transfer type, focusing on transferring patients from the scene to the hospital, with the operational mode being "hospital take-off / landing point—demand point—tertiary hospital—original hospital." This scenario does not restrict the return to the original take-off point, allowing for more flexible route selection.

[0082] Figure 3 The diagram shown illustrates the site selection layout results for Scenario 1 constructed in this embodiment. The diagram spatially displays the relationship between candidate take-off and landing points, demand points, and the final site selection results. Gray hollow circles represent candidate take-off and landing points, blue squares represent medical emergency demand points, and red solid circles represent take-off and landing points selected by the optimization model. Blue lines represent the coverage relationship between each demand point and its corresponding service take-off and landing point.

[0083] Figure 4 The diagram shown illustrates the site selection layout results for Scenario 2 constructed in this embodiment. The diagram shows that in Scenario 2, eVTOL executes a transfer task from "hospital take-off / landing point—demand point—tertiary hospital—original hospital," without returning to the original take-off / landing point. Therefore, the path organization is more flexible, and the site selection layout places greater emphasis on transfer efficiency. Figure 4 a and Figure 4As can be seen from b, the red take-off and landing points exhibit a relatively balanced spatial distribution, covering both the central high-demand area and the peripheral scattered demand points, reflecting a comprehensive optimization of the overall service range and transit accessibility.

[0084] S4. Construct the Maximum Coverage Location Optimization Model (MCLP) considering multiple constraints. Establish the maximum coverage location optimization model with the objective of maximizing the demand size effectively covered. Objective function:

[0085]

[0086] Where Z represents the maximum number of coverage demand points; I represents the final set of rescue demand points i; J represents the set of candidate take-off and landing points; w i This represents the weight of the medical urgency level at the final rescue demand point i; y ij This indicates whether the final rescue request point i is served by the candidate take-off and landing point j.

[0087] Constraints include the number of takeoff and landing sites, maximum response time, maximum eVTOL range, and maximum service capacity per site (to prevent single-point overload). Coverage determination: Flight time and range are calculated based on the urban air medical emergency rescue operation scenario defined in step S3. If the time is less than the threshold, it is determined to be effective coverage.

[0088] (1) Constraints on the number of take-off and landing points

[0089]

[0090] Where j represents a candidate take-off and landing point; J represents the set of candidate take-off and landing points; x j This indicates whether candidate take-off and landing point j has been selected for construction; P represents the maximum number of take-off and landing points allowed to be constructed during the planning period.

[0091] (2) Service feasibility constraints

[0092]

[0093] in, This indicates that the round-trip flight time from "hospital take-off / landing point → demand point i → return to the original hospital" meets the response time and range constraints; otherwise... .

[0094] (3) Unique service constraint

[0095]

[0096] (4) Single station capacity constraints

[0097]

[0098] Where K represents the maximum service capacity of a single take-off and landing point.

[0099] (5) Variable value constraints

[0100]

[0101] Scenario 2 shares the same planning objective as Scenario 1: to maximize the scale of emergency rescue needs effectively covered, given a limited number of landing and takeoff points. However, their service feasibility constraints differ slightly. Scenario 2's service feasibility constraints are as follows:

[0102]

[0103] in, This means that the total flight time from the starting hospital, through the final rescue demand point i, and to the nearest available hospital to the final rescue demand point i meets the response time and range constraints; otherwise... .

[0104] S5. An improved genetic algorithm is used to solve the model. For the constructed site selection optimization model, an improved genetic algorithm is introduced to solve it. The computational complexity is reduced by preprocessing the coverage relationship, and the feasibility of the solution is ensured by combining constraint embedding and solution repair mechanism. Under the premise of satisfying multiple constraints, the efficient optimization solution of the take-off and landing point layout scheme is achieved.

[0105] S51. Considering the characteristics of the location selection problem, a genetic algorithm (GA) is designed to solve it. Binary encoding is used to represent the location status of takeoff and landing points, and the fitness function is set to the total coverage requirement. The algorithm iteratively searches through selection, crossover, and mutation operations until the termination condition is met, outputting the optimal location solution.

[0106] For the maximum coverage location problem, a fitness function is constructed with the objective of maximizing the total coverage demand, where:

[0107] in, This indicates whether, under time and range constraints, candidate take-off and landing point j can serve the final rescue demand point i.

[0108] The overriding variable is defined as:

[0109]

[0110] Among them, y i The coverage state variable represents the final rescue demand point i. That is, when there is at least one selected take-off and landing point that can serve the final rescue demand point i under the constraints, the final rescue demand point is considered to be covered.

[0111] The fitness function is defined as:

[0112]

[0113] Where F represents the individual fitness value.

[0114] Simultaneously satisfying the constraint on the number of take-off and landing points:

[0115]

[0116] When an individual does not meet the above constraints, its fitness is assigned a minimum value. This is to reduce the probability of it being selected.

[0117] Where the covering relationship a ij The time and range constraints have been predetermined:

[0118]

[0119] Among them, T ij D represents the time between the final rescue demand point i and the candidate take-off and landing point j; ij T represents the flight distance from candidate takeoff and landing point j to the final rescue demand point i; max δ represents the maximum permissible response time threshold; δ represents the distance threshold between the final rescue demand point i and the candidate take-off and landing point j.

[0120] S52. Selection Operator: A roulette wheel selection method based on fitness ratio is adopted. After calculating the fitness of individuals in the population, high-quality individuals need to be selected from the current population to enter the next generation, so as to achieve gradual optimization of the solution. The core role of the selection operator is to balance population diversity with the principle of survival of the fittest, that is, to increase the probability of high-fitness individuals being retained and replicated, while avoiding local optima caused by premature convergence. For the take-off and landing point location optimization model constructed in this paper, its fitness function can reflect the coverage effect and service performance of the layout scheme. Therefore, a roulette wheel selection method based on fitness ratio is adopted, so that the probability of an individual being selected is proportional to its fitness value. This preserves high-quality solutions while providing some evolutionary opportunities for suboptimal solutions, thereby enhancing the global search capability of the algorithm. The specific steps are as follows:

[0121] ① The fitness of each individual in the population is nonnegated:

[0122]

[0123] in, This represents the initial fitness value of the k-th individual; Indicates the first The nonnegation fitness value of an individual; min(f) represents the minimum fitness value of all individuals in the current population.

[0124] ② Calculate the individual's choice probability:

[0125]

[0126] in, represents the probability that the k-th individual is selected; N represents the population size.

[0127] ③ Construct the cumulative probability distribution:

[0128]

[0129] in, p represents the cumulative selection probability corresponding to the k-th individual; t Let represent the selection probability of the t-th individual; t represents the individual index.

[0130] ④ Generate a random number r within the interval [0,1], and select one that satisfies: Individuals enter the next generation of the population.

[0131] S53. Single-point crossover is adopted, combined with a solution repair mechanism to ensure constraint satisfaction. After the selection operation, although high-quality individuals are retained, the overall solution space structure of the population does not change substantially. Therefore, it is necessary to use the crossover operator to realize the information recombination between individuals to generate new potential optimal solutions. The crossover operation, by combining gene fragments of different individuals, helps to explore new feasible regions in the solution space and is a key mechanism for the genetic algorithm to achieve global search capability. For the take-off and landing point selection problem studied in this paper, chromosomes usually use 0-1 encoding to represent the selection state of candidate points. The single-point crossover method can achieve effective recombination between different layout schemes while maintaining the simplicity of the encoding structure. However, since the crossover operation may destroy the original individuals' satisfaction of the constraint on the number of take-off and landing points, it is necessary to introduce a solution repair mechanism after crossover to ensure that the generated individuals always satisfy the facility quantity constraint, thereby ensuring that the algorithm search process always takes place within the feasible solution space. The specific steps are as follows:

[0132] ① Randomly select parent individuals based on crossover probability Pc; ② Randomly select crossover point c within the chromosome length range; ③ Exchange gene segments after the crossover point between the two individuals to generate offspring individuals; ④ Perform feasibility repair on the individuals after crossover:

[0133] like Randomly set any extra "1"s to "0"s;

[0134] like Randomly set some "0"s to "1"s until the constraint is satisfied.

[0135] S6. Generate a hierarchical take-off and landing point layout scheme and service network. Based on the optimal location result, the take-off and landing point layout result analysis method analyzes the spatial distribution characteristics and overall coverage effect of the take-off and landing points to achieve key node identification and layout optimization.

[0136] First, based on the optimal location results output by the optimization model, the selected set of take-off and landing points is extracted, and combined with their spatial coordinate information, the overall distribution characteristics of the take-off and landing points are analyzed to identify their spatial configuration in high-demand areas and peripheral areas.

[0137] Secondly, based on the pre-constructed coverage relationship matrix, under the conditions of meeting response time and range constraints, the coverage of each take-off and landing point to the demand points is analyzed, and the overall coverage status of all selected take-off and landing points to the demand points is further statistically analyzed to evaluate the coverage effect of the layout scheme, including the number of covered demand points and the distribution of uncovered demand points.

[0138] Based on this, and combined with the coverage results, we identify areas of demand that are not covered or are under-covered, analyze the service gaps in the spatial layout of existing take-off and landing points, and thus reveal the shortcomings of the layout scheme in local areas.

[0139] Finally, based on the role of take-off and landing points in the overall coverage system and their spatial location characteristics, they are divided into different types, including key nodes that play a core coverage role in the main demand areas, and auxiliary nodes used to make up for insufficient coverage in edge areas or local areas, thus forming a structured take-off and landing point layout scheme.

[0140] Based on the experimental results, a differentiated eVTOL emergency medical network with 15 points as the critical scale was constructed for XX City. Through multi-scenario iterative simulation, the optimal layout scheme was determined to consist of 15 take-off and landing points with different functions: nodes 73, 100, 106, and 74 serve as the core hub layer, ensuring 100% basic coverage robustness across the entire area; nodes 28, 30, and 89 serve as the medical station coupling layer, improving the transport mode coverage rate to the experimental optimal solution (73.17%); supplemented by elastic supplementary point layers such as nodes 1, 11, and 22, dynamic coverage balance is achieved when the response threshold shrinks to 10 minutes. This layout achieves the spatially optimal configuration of system rescue efficiency under the dual constraints of resource constraints (maximum service capacity of 5 per station) and time constraints.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for layout of urban aviation emergency medical take-off and landing points for eVTOL considering demand point clustering, characterized in that, The method includes the following steps: S1. Using urban point of interest data as a spatial proxy variable for potential medical emergency needs, select functional types related to the occurrence of medical emergency events as raw data, clean, deduplicate and transform the data to construct an initial set of potential medical emergency needs points. S2. A density-based spatial clustering algorithm is used to process the initial potential medical emergency demand points. A distance metric method suitable for geospatial data is introduced to calculate the distance relationship between points. By setting the neighborhood radius and minimum sample number parameters, the initial potential medical emergency demand points in space are clustered, and the discretely distributed initial potential medical emergency demand points are aggregated into several demand clusters. The center point of each demand cluster is extracted as the final rescue demand point, while noise points are removed. S3. Based on the essential differences in the service objectives, time constraints, and operational paths of eVTOL urban medical emergency rescue missions, eVTOL urban medical emergency rescue missions are divided into two typical scenarios: direct medical rescue and indirect medical transport. S4. With the goal of maximizing the effective coverage of medical emergency needs, a take-off and landing point location model is constructed. The model comprehensively considers the limitations on the number of take-off and landing points, flight response time constraints, flight range constraints, and service capacity constraints of a single take-off and landing point. The coverage relationship between the final rescue demand point and the candidate take-off and landing points is established, and a complete mathematical programming model is formed to characterize the relationship between the layout of take-off and landing points and the demand coverage. S5. For the location selection model constructed in S4, an improved genetic algorithm is introduced to solve it. The computational complexity is reduced by preprocessing the coverage relationship, and the feasibility of the solution is ensured by combining constraint embedding and solution repair mechanism. Under the premise of satisfying the constraint conditions, the solution of the take-off and landing point layout scheme is realized. S6. Analysis method of take-off and landing point layout results based on site selection results: By analyzing the spatial distribution characteristics and overall coverage effect of take-off and landing points, key nodes can be identified and laid out.

2. The method for eVTOL urban aviation emergency medical take-off and landing point layout considering demand point clustering as described in claim 1, characterized in that, In step S2, the neighborhood radius ɛ is defined as: Among them, s m N represents the m-th initial potential medical emergency demand point; ɛ (s m ) represents all pairs of s m The set of sample points whose distance does not exceed ɛ; D represents the initial set of potential medical emergency demand points; dist(·) represents the spatial distance between the initial potential medical emergency demand points, s q This represents any initial potential medical emergency demand point in the dataset; The method for determining the neighborhood radius ɛ is as follows: for each initial potential medical emergency demand point, calculate its distance to the k-th nearest neighbor point, arrange the distance values ​​of all points in ascending order, and plot a k-distance curve. The distance value corresponding to the inflection point where the curve changes from a gradual change to a steep increase is used as the reference value of the neighborhood radius ɛ. The minimum sample size parameter is determined based on the comprehensive data scale and spatial distribution characteristics.

3. The method for eVTOL urban aviation emergency medical take-off and landing point layout considering demand point clustering as described in claim 1, characterized in that, In step S3, the direct medical rescue scenario is responsive, focusing on the delivery of emergency supplies or the arrival of emergency personnel. The operating mode is hospital take-off and landing point—demand point—return to the original hospital, and the constraint is that the round-trip flight time and distance must meet the threshold. The indirect medical transport scenario is transport-oriented, focusing on transferring patients from the scene to the hospital. The operating mode is hospital take-off and landing point—demand point—tertiary hospital—original hospital. This scenario does not limit the return to the original take-off point.

4. The method for eVTOL urban aviation emergency medical take-off and landing point layout considering demand point clustering as described in claim 1, characterized in that, In step S4, the objective function of the take-off and landing point selection model is: Where Z represents the maximum number of coverage demand points; I represents the final set of rescue demand points i; J represents the set of candidate take-off and landing points; w i This represents the weight of the medical urgency level at the final rescue demand point i; y ij This indicates whether the final rescue request point i is served by the candidate take-off and landing point j.

5. The method for eVTOL urban aviation emergency medical take-off and landing point layout considering demand point clustering as described in claim 1, characterized in that, The specific steps in step S5 are as follows: S51. To address the characteristics of the location selection problem, a genetic algorithm (GA) is designed to solve it. The binary encoding is used to represent the location status of the take-off and landing points. The fitness function is designed to cover the total demand. The algorithm iteratively searches through selection, crossover, and mutation operations until the termination condition is met, and then outputs the location selection scheme. The fitness function is defined as: Where F represents the individual fitness value; S52. Selection operator: Using a roulette wheel selection method based on fitness ratios, after calculating the fitness of individuals in the population, individuals need to be selected from the current population to enter the next generation. ① The fitness of each individual in the population is nonnegated: in, This represents the initial fitness value of the k-th individual; Indicates the first The nonnegated fitness value of each individual; min(f) represents the minimum fitness value of all individuals in the current population; ② Calculate the individual's choice probability: in, This represents the probability that the k-th individual is selected; N represents the population size. ③ Construct the cumulative probability distribution: Where, q k p represents the cumulative selection probability corresponding to the k-th individual; t This represents the selection probability of the t-th individual; t represents the individual index. ④ Generate a random number r within the interval [0,1], and select one that satisfies: Individuals enter the next generation of the population; S53. Single-point crossover is adopted, combined with a solution repair mechanism, to ensure that the generated individuals always meet the facility quantity constraint and to ensure that the algorithm search process always takes place within the feasible solution space. The specific steps are as follows: ① Randomly select parent individuals based on crossover probability Pc; ② Randomly select crossover point c within the chromosome length range; ③ Exchange gene segments after the crossover point between the two individuals to generate offspring individuals; ④ Perform feasibility repair on the individuals after crossover: like Randomly set any extra "1"s to "0"s; like Randomly set some "0"s to "1"s until the constraint is satisfied; Where, x j This indicates whether candidate take-off and landing point j has been selected for construction; P represents the maximum number of take-off and landing points allowed to be constructed during the planning period.

Images