Low-altitude medical rescue aircraft site selection and scheduling optimization method based on adaptive large neighborhood search

Abstract

The invention discloses a low-altitude medical rescue aircraft site selection and scheduling optimization method based on adaptive large neighborhood search, and the method comprises the steps: obtaining a weight evaluation result of each medical rescue demand point according to the multi-dimensional basic data of an urban region; according to the demand point weight evaluation result, constructing a site selection-demand allocation mixed integer programming model; according to the mixed integer programming model, obtaining an initial feasible solution through K-means clustering and a greedy heuristic algorithm; according to the initial feasible solution, obtaining an optimal solution through iterative operation of a destruction operator and a repair operator and local search of a variable neighborhood descent method; and according to the optimal solution, obtaining a temporary take-off and landing point layout scheme and an emergency rescue scheduling scheme. According to the method, the purposes of minimizing global rescue response time and realizing fair and efficient coverage of medical rescue resources are achieved through scientific site selection and reasonable configuration of rescue resources of temporary take-off and landing points.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of urban low-altitude traffic and medical emergency rescue, and particularly relates to a low-altitude medical rescue aircraft site selection and scheduling optimization method based on adaptive large neighborhood search. BACKGROUND

[0002] With the rapid development of low-altitude economy, the urban low-altitude traffic system shows broad application prospects in the fields of emergency rescue, medical transfer, disaster emergency and logistics distribution. Low-altitude medical rescue, as an important application scenario of low-altitude economy, has high dependence on the completeness of low-altitude infrastructure and the efficiency of air-ground collaborative scheduling. However, the low-altitude infrastructure system in China is not perfect at present, and problems such as insufficient number of temporary landing points, unbalanced spatial layout, and complex restrictions of urban no-fly zones have become key bottlenecks restricting the landing of low-altitude economy and the improvement of emergency medical rescue efficiency. The existing low-altitude medical rescue system mainly depends on existing fixed airports or the parking lots of large medical institutions, and its layout is limited by urban construction conditions and air control, so it is difficult to achieve rapid response and coverage optimization in emergency situations. At the same time, traditional site selection and scheduling methods are usually modeled independently, and factors such as ground transportation time, aircraft range limit, no-fly zone constraint and collaborative scheduling mechanism are not considered comprehensively, resulting in insufficient rescue efficiency and low resource utilization. Therefore, under the background of imperfect low-altitude infrastructure, how to achieve fair and efficient coverage of medical rescue resources, shorten the global rescue response time, and improve the overall resilience of the urban low-altitude emergency system through scientific temporary landing site selection and resource allocation decisions has become a technical problem to be solved in this field. SUMMARY

[0003] To solve the above technical problems, the application provides a low-altitude medical rescue aircraft site selection and scheduling optimization method based on adaptive large neighborhood search, which can minimize the global rescue response time and achieve fair and efficient coverage of medical rescue resources through scientific site selection of temporary landing points and reasonable allocation of rescue resources.

[0004] To achieve the above purpose, the application provides a low-altitude medical rescue aircraft site selection and scheduling optimization method based on adaptive large neighborhood search, which comprises the following steps:

[0005] According to the multi-dimensional basic data of the urban area, the weight evaluation results of each medical rescue demand point are obtained;

[0006] According to the demand point weight evaluation results, a site selection-demand allocation mixed integer programming model is constructed;

[0007] According to the mixed integer programming model, an initial feasible solution is obtained through K-means clustering and greedy heuristic algorithm;

[0008] According to the initial feasible solution, an optimal solution is obtained through iterative operation of a destruction operator and a repair operator and local search of a variable neighborhood descent method;

[0009] According to the optimal solution, a temporary take-off and landing point layout scheme and an emergency rescue scheduling scheme are obtained.

[0010] Optionally, according to multi-dimensional basic data of a city area, a weight evaluation result of each medical rescue demand point is obtained, including:

[0011] According to population density, number of old people, take-off safety coefficient, obstacle density, medical accessibility distance, GDP level and traffic convenience index data, a standardized data matrix after positive direction processing of the indexes is obtained;

[0012] According to the standardized data matrix, information entropy values and redundancy of each index are calculated through an entropy weight method, and objective weight values of each index are obtained;

[0013] According to the objective weight values of each index and the standardized data matrix, a comprehensive score value of each demand point is obtained through linear weighted summation;

[0014] According to the comprehensive score value, the weight evaluation result is obtained.

[0015] Optionally, according to the comprehensive score value, the weight evaluation result is obtained, including:

[0016] The comprehensive score value is arranged in descending order;

[0017] The arranged comprehensive score is divided into a preset number of grades, and corresponding weight values are assigned, and the weight evaluation result is obtained.

[0018] Optionally, according to the demand point weight evaluation result, a site-demand allocation mixed integer programming model is constructed, including:

[0019] According to a candidate take-off and landing point set, a demand point set, a hospital set, aircraft performance parameters and no-fly zone data, a decision variable system is obtained;

[0020] According to the decision variable system and rescue response time calculation rules of each demand point, a target function is obtained;

[0021] According to the evaluation result in the demand point, take-off and landing point construction cost, demand point allocation relationship, aircraft scheduling rules, no-fly zone safety distance and aircraft range limit, a constraint condition set is obtained;

[0022] According to the target function and the constraint condition set, the site-demand allocation mixed integer programming model is obtained.

[0023] Optionally, the rescue response time calculation rule of each demand point comprises:

[0024] For the demand point using direct ground transfer, the rescue response time is the ground transportation time from the demand point to the target point;

[0025] For the demand point using air transfer, the rescue response time is determined by the larger value of the ground time from the demand point to the temporary take-off and landing point and the flight time of the aircraft from the airport to the temporary take-off and landing point, and the flight time from the temporary take-off and landing point to the target point.

[0026] Optionally, according to the mixed integer programming model, the initial feasible solution is obtained by K-means clustering and greedy heuristic algorithm, comprising:

[0027] According to the mixed integer programming model, the K-means clustering is used to determine the take-off and landing point;

[0028] According to the take-off and landing point, the greedy heuristic algorithm is used to determine the shortest rescue time path for each demand point to obtain the initial feasible solution.

[0029] Optionally, according to the initial feasible solution, the optimal solution is obtained by iterative operation of the destruction operator and the repair operator and local search of the variable neighborhood descent method, comprising:

[0030] The destruction operator is used to destroy the initial feasible solution to obtain a destroyed solution, wherein the destruction operator comprises a random destruction operator and a bottleneck-oriented destruction operator;

[0031] According to the destroyed solution, the unallocated demand points are re-allocated by the repair operator to obtain a repaired solution, wherein the repair operator comprises a greedy repair operator and a k-remorse repair operator;

[0032] Based on the repaired solution, the local search of the variable neighborhood descent method is combined to obtain the optimal solution.

[0033] Optionally, the initial feasible solution is destroyed by the bottleneck-oriented destruction operator to obtain a destroyed solution, comprising:

[0034] According to the number of service demands and the average arrival time of each take-off and landing point in the initial feasible solution, a take-off and landing point performance ranking result is obtained;

[0035] According to the ground transportation time, flight time and total rescue time of each demand point in the initial feasible solution, a bottleneck type identification result is obtained;

[0036] According to the performance ranking result and the bottleneck type identification result, the take-off and landing points with poor performance or bottleneck demand points are preferentially removed to obtain the destroyed solution.

[0037] Optionally, according to the destruction solution, the unallocated demand points are re-allocated by the k-remedy repair operator to obtain a repair solution, which comprises:

[0038] According to the rescue time of each unallocated demand point to each take-off and landing point and hospital, the regret value of k optimal solutions is obtained;

[0039] According to the sorting result of the regret value from large to small, the demand point with the largest regret value is preferentially allocated to the rescue scheme to obtain the repair solution.

[0040] Optionally, obtaining the optimal solution further comprises:

[0041] According to the historical success times and use times of each destruction operator and repair operator, an operator weight update value is obtained;

[0042] According to the operator weight update value, the destruction operator and repair operator used in the next iteration period are obtained through a roulette wheel selection mechanism.

[0043] Compared with the prior art, the present application has the following advantages and technical effects:

[0044] The present application can objectively quantify the medical service demand intensity of each potential demand point through the entropy weight method, avoid subjective weighting bias, and provide a scientific demand analysis basis for subsequent site selection decision-making; can comprehensively consider rescue time efficiency, construction cost, flight restricted area constraints, aircraft range restrictions and other multi-dimensional constraint conditions through a site selection-demand allocation mixed integer programming model, realize joint optimization decision of temporary take-off and landing point layout position, demand point medical treatment mode, aircraft flight route and demand point and take-off and landing point allocation relationship; can efficiently solve large-scale site selection and scheduling problems through the EALNS algorithm combined with K-means clustering, destruction-repair operator, variable neighborhood descent method and elite pool strategy, avoid memory overflow, long solving time and other problems of commercial solvers in large-scale examples; can output the optimal layout scheme of temporary take-off and landing points, the rescue scheduling scheme of each demand point and the system performance index based on the solving result, and provide a quantitative decision basis for urban low-altitude medical rescue system planning. The present application realizes the purpose of minimizing global rescue response time and realizing fair and efficient coverage of medical rescue resources by scientifically selecting temporary take-off and landing points and reasonably configuring rescue resources. BRIEF DESCRIPTION OF DRAWINGS

[0045] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The embodiments of this application and its description together with the drawings make abasis of the present application. In the drawings:

[0046] Figure 1 is a low-altitude medical rescue aircraft site selection and scheduling optimization method flowchart based on adaptive large neighborhood search. Detailed Implementation

[0047] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0049] Explanation of the English abbreviations and their meanings involved in this embodiment:

[0050] LRP (Location-Routing Problem): A combinatorial optimization problem that simultaneously optimizes facility location and route planning. EALNS (Enhanced Adaptive Large Neighborhood Search): An enhanced adaptive large neighborhood search algorithm, a metaheuristic optimization algorithm. ALNS (Adaptive Large Neighborhood Search): An adaptive large neighborhood search algorithm, the foundational framework of EALNS. VND (Variable Neighborhood Descent): A local search algorithm. K-means: A K-means clustering algorithm, a classic unsupervised learning algorithm. EWM (Entropy Weight Method): An entropy weighting method, an objective weighting method based on information theory. GDP (Gross Domestic Product): Used to measure the level of regional economic development. Jaccard Coefficient: A statistical measure of the similarity between two sets, calculated as the size of the intersection of the two sets divided by the size of the union; in this example, it is used to quantify the similarity between two solutions.

[0051] This embodiment proposes a method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search, such as... Figure 1 As shown, the specific steps include:

[0052] Based on multi-dimensional basic data of urban areas, obtain the weighted evaluation results of each medical rescue demand point;

[0053] Based on the evaluation results of demand point weights, a hybrid integer programming model for site selection and demand allocation is constructed.

[0054] According to the mixed integer programming model, an initial feasible solution is obtained through K-means clustering and a greedy heuristic algorithm;

[0055] According to the initial feasible solution, an optimal solution is obtained through the iterative operation of the destruction operator and the repair operator and the local search of the variable neighborhood descent method;

[0056] According to the optimal solution, a temporary take-off and landing point layout scheme and an emergency rescue scheduling scheme are obtained.

[0057] Specifically, the city area is discretized by gridding, and the medical service intensity of each demand point is quantified based on population quantity and spatial distribution data. A site selection-demand allocation mixed integer programming model is established to minimize the global rescue response time, and the actual operation constraints such as take-off and landing point construction cost, urban no-fly zone airspace constraint, aircraft range limit and air-ground cooperation mechanism are comprehensively considered to jointly determine the layout location of temporary take-off and landing point, the selection of medical transportation mode of each demand point and the allocation relationship between demand point and temporary take-off and landing point. For large-scale problem solving, an enhanced adaptive large neighborhood search algorithm is designed to generate an initial feasible solution through K-means clustering and greedy heuristic, and a multi-type destruction-repair operator is used to explore the solution space under the adaptive large neighborhood search framework and embedded variable neighborhood descent method to implement local search. The elite solution update and operator weight adaptive adjustment mechanism are used to improve the solving efficiency. The algorithm performance is verified through empirical examples, and the influence law of key parameters such as ground traffic speed and take-off and landing point construction cost on system rescue time and layout scheme is tested through sensitivity analysis.

[0058] More specifically, based on a multi-dimensional index system and an entropy weight method, the potential medical rescue demand points in the urban area are evaluated by weight;

[0059] A site selection-demand allocation mixed integer programming model is established to minimize the global rescue response time, and the layout location of temporary take-off and landing point, the selection of medical transportation mode of demand point, the flight route of aircraft under transfer and the allocation relationship between demand point and take-off and landing point are comprehensively determined;

[0060] An enhanced adaptive large neighborhood search algorithm is designed to solve the optimization model, an initial solution is generated through K-means clustering and greedy heuristic, a multi-type destruction-repair operator is used to iteratively update the solution, and a variable neighborhood descent method is embedded for local improvement;

[0061] Based on the solving result, an optimal layout scheme of temporary take-off and landing point and a rescue scheduling scheme of each demand point are obtained.

[0062] Further, based on the multi-dimensional basic data of the urban area, the weight evaluation results of each medical rescue demand point are obtained, including:

[0063] According to the population density, the number of the elderly population, the take-off safety coefficient, the obstacle density, the medical accessibility distance, the GDP level and the traffic convenience index data, a standardized data matrix after positive treatment of the indexes is obtained;

[0064] According to the standardized data matrix, the information entropy value and the redundancy of each index are calculated by the entropy weight method, and the objective weight value of each index is obtained;

[0065] According to the objective weight value of each index and the standardized data matrix, the comprehensive score value of each demand point is obtained by linear weighted summation;

[0066] According to the comprehensive score value, the weight evaluation result is obtained.

[0067] Specifically, the method for weight evaluation of potential medical rescue demand points comprises:

[0068] A multi-dimensional medical demand index system is constructed, and seven indexes including total population density, number of elderly population, take-off safety coefficient, obstacle density, medical accessibility distance, GDP level and traffic convenience are selected as evaluation factors. All negative indexes are converted into positive indexes by taking the reciprocal;

[0069] The entropy weight method is used to calculate the objective weight of each index, the original data is processed by minimum-maximum standardization, the index proportion matrix is constructed, the information entropy of each index is calculated, and the index weight is calculated by redundancy;

[0070] The linear weighted summation of the standardized value of each index and the corresponding weight is performed to obtain the comprehensive evaluation score of each demand point. According to the comprehensive score, the demand points are sorted and divided into five levels, and the corresponding weight values are given as the quantitative parameters of the importance of the demand points in the subsequent site selection optimization model.

[0071] Further, the weight evaluation result comprises:

[0072] The comprehensive score values are arranged in descending order;

[0073] The arranged comprehensive score is divided into a preset number of levels, and the corresponding weight values are given to obtain the weight evaluation result.

[0074] Further, according to the weight evaluation result of the demand points, a site-demand allocation mixed integer programming model is constructed, which comprises:

[0075] According to the candidate take-off and landing point set, the demand point set, the hospital set, the aircraft performance parameters and the no-fly zone data, a decision variable system is obtained;

[0076] According to the decision variable system and the rescue response time calculation rules of each demand point, a target function is obtained;

[0077] According to the construction cost of the take-off and landing point, the demand point distribution relationship, the aircraft scheduling rule, the safety distance of the no-fly zone and the aircraft range limit, a constraint condition set is obtained;

[0078] According to the objective function and the constraint condition set, a site-demand distribution mixed integer programming model is obtained.

[0079] Specifically, the method for establishing the site-demand distribution mixed integer programming model comprises:

[0080] Defining a decision variable system, including temporary take-off and landing point site selection variables, demand point and take-off and landing point distribution variables, demand point medical treatment mode variables, demand point and hospital distribution variables, aircraft scheduling variables and auxiliary decision variables;

[0081] Constructing an objective function, taking the minimum value of the maximum rescue response time of all demand points in the rescue network as the optimization target, for the demand points using direct ground transfer, the rescue response time is the ground transportation time from the demand point to the hospital, for the demand points using air transfer, the rescue response time is jointly determined by the larger value of the ground time from the demand point to the temporary take-off and landing point and the flight time from the airport to the temporary take-off and landing point, and the flight time from the temporary take-off and landing point to the hospital;

[0082] Establishing a constraint condition system, including temporary take-off and landing point site selection cost constraints, demand point distribution constraints, hospital distribution constraints, aircraft scheduling constraints, no-fly zone constraints, aircraft range constraints and response time constraints;

[0083] Linearizing the nonlinear constraints, introducing auxiliary variables to represent the time point when the demand point is ready to go to the hospital from the temporary take-off and landing point, and introducing auxiliary decision variables to realize linear representation of the product of multiple decision variables.

[0084] Further, the rescue response time calculation rule of each demand point comprises:

[0085] For demand points using direct ground transfer, the rescue response time is the ground transportation time from the demand point to the target point;

[0086] For demand points using air transfer, the rescue response time is jointly determined by the larger value of the ground time from the demand point to the temporary take-off and landing point and the flight time from the airport to the temporary take-off and landing point, and the flight time from the temporary take-off and landing point to the target point.

[0087] Further, according to the mixed integer programming model, an initial feasible solution is obtained by K-means clustering and greedy heuristic algorithm, comprising:

[0088] According to the mixed integer programming model, the K-means clustering is used to determine the take-off and landing point;

[0089] According to the take-off and landing points, a greedy heuristic algorithm is used to determine the shortest rescue time path for each demand point to obtain an initial feasible solution.

[0090] Further, according to the initial feasible solution, an optimal solution is obtained through iterative operations of a destruction operator and a repair operator and local search of a variable neighborhood descent method, and the optimal solution comprises the following steps:

[0091] The initial feasible solution is destroyed by using a destruction operator to obtain a destroyed solution, wherein the destruction operator comprises a random destruction operator and a bottleneck-oriented destruction operator;

[0092] According to the destroyed solution, the unallocated demand points are re-allocated by using a repair operator to obtain a repaired solution, wherein the repair operator comprises a greedy repair operator and a k-regret repair operator;

[0093] Based on the repaired solution, an optimal solution is obtained by combining local search of a variable neighborhood descent method.

[0094] Further, the initial feasible solution is destroyed by using a bottleneck-oriented destruction operator to obtain a destroyed solution, and the method comprises the following steps:

[0095] According to the service demand quantity and the average arrival time of each take-off and landing point in the initial feasible solution, a performance ranking result of the take-off and landing points is obtained;

[0096] According to the ground transportation time, the flight time and the total rescue time of each demand point in the initial feasible solution, a bottleneck type identification result is obtained;

[0097] According to the performance ranking result and the bottleneck type identification result, a take-off and landing point with poor performance or a bottleneck demand point is preferentially removed to obtain a destroyed solution.

[0098] Further, according to the destroyed solution, the unallocated demand points are re-allocated by using a k-regret repair operator to obtain a repaired solution, and the method comprises the following steps:

[0099] According to the rescue time of each unallocated demand point to each take-off and landing point and hospital, a regret value of k optimal schemes is obtained;

[0100] According to the ranking result of the regret values from large to small, a demand point with the largest regret value is preferentially allocated with a rescue scheme to obtain a repaired solution.

[0101] Further, the optimal solution further comprises the following steps:

[0102] According to the historical success times and the use times of each destruction operator and repair operator, an operator weight update value is obtained;

[0103] According to the operator weight update value, a destruction operator and a repair operator used in a next iteration period are obtained through a roulette wheel selection mechanism.

[0104] Specifically, the methods for designing the EALNS algorithm to solve optimization models include:

[0105] Define the representation structure of the solution. The complete solution consists of two parts: a temporary take-off and landing point location scheme and a demand point allocation scheme. Each demand point corresponds to a five-tuple allocation scheme.

[0106] To generate an initial feasible solution, a K-means clustering algorithm with a greedy strategy is used. First, at the temporary take-off and landing point selection level, the K-means clustering algorithm is run to identify the spatial distribution pattern of the demand points and map it to the actual candidate take-off and landing points. Then, at the demand point allocation level, a greedy heuristic algorithm is run to determine the path with the shortest rescue time for each demand point.

[0107] The design includes a library of destruction operators, including random destruction operators and bottleneck-oriented destruction operators. Random destruction operators randomly select temporary take-off and landing points and demand point allocation schemes for destruction, while bottleneck-oriented destruction operators implement targeted destruction by analyzing the bottleneck causes of the solution.

[0108] The design includes a repair operator library, which includes a greedy repair operator and a k-regret repair operator. The greedy repair operator selects the optimal rescue plan for each unassigned demand under the current state, while the k-regret repair operator prioritizes the demand with the largest regret value by calculating the response time difference between the top k optimal choices.

[0109] The design incorporates an operator selection and weight update mechanism. Initially, all destruction and repair operators have the same weight. Operators are selected based on the roulette wheel principle, and their weights are dynamically adjusted after each iteration based on the historical performance of the operators.

[0110] The neighborhood descent method with embedded variables is used for local search. The VND algorithm is periodically embedded in the main loop of ALNS to optimize the current solution locally. Four neighborhood structures are used in sequence for search: take-off and landing point fine-tuning, hospital selection optimization, aircraft configuration optimization and transportation mode switching.

[0111] The simulated annealing acceptance criterion is adopted, always accepting the better new solution and accepting the worse new solution with a certain probability.

[0112] Maintain an elite solution pool to store high-quality solutions with different structures, dynamically adjust diversity requirements based on the degree of stagnation, and determine the similarity of solutions through Jaccard system quantitative methods.

[0113] Set the algorithm termination condition. The algorithm will terminate and output the optimal solution when the specified maximum number of iterations is reached or the solution reaches a specified number of consecutive improvements.

[0114] The solution results yielded the optimal layout scheme for temporary take-off and landing points and the rescue dispatch scheme for each demand point, including:

[0115] The temporary take-off and landing point layout plan determines which candidate locations within the urban area should be used to construct temporary take-off and landing points, and each selected take-off and landing point corresponds to its geographical coordinates, construction cost, and service area.

[0116] The demand point rescue dispatch plan determines the rescue method, corresponding hospital, and temporary take-off and landing point, departure airport and aircraft type for each demand point in the case of transfer;

[0117] The system performance metrics calculate and output key performance indicators of the rescue network, including maximum rescue response time, average rescue response time, and the distribution of rescue time at each demand point.

[0118] This embodiment takes a certain city as the research object, and the specific implementation steps are as follows:

[0119] (I) Weighting method for potential medical rescue points based on entropy weight method:

[0120] Step 1: Construct a multi-dimensional evaluation index system:

[0121] Basic data of streets in various administrative districts of a city were collected to construct seven evaluation indicators:

[0122] (1) Total population density The calculation formula is: (Unit: people / km²)

[0123] In the formula, The total population of the street (in people); This refers to the street area (km²). For example, given the total population of a certain street... =120,000 people, street area =15km², which can be obtained from the above formula. =8000 people / km².

[0124] (2) Number of elderly people The population statistics for those aged 60 and above are used (unit: person).

[0125] (3) Takeoff safety factor The calculation formula is: ;

[0126] In the formula, The height (m) of the largest building within a 500m radius; The safety threshold is 100m. For example, ,but =0.65.

[0127] (4) Obstacle density (Negative indicator), the calculation formula is: (Unit: units / km²);

[0128] In the formula, This refers to the number of buildings.

[0129] (5) Medical accessibility distance (Negative index), using the Haversine formula: ;

[0130] In the formula, =6371km; Δlat and Δlon are the difference in latitude and longitude (in radians).

[0131] (6) GDP level The data used is from the statistical yearbook (in 100 million yuan).

[0132] (7) Convenience of transportation The calculation formula is: ;

[0133] In the formula, The number of bus stops; This refers to the number of subway stations.

[0134] Step 2: Indicator Standardization:

[0135] Turning negative indicators positive: , .

[0136] Minimum-maximum standardization is adopted: ;

[0137] For example, a certain street .

[0138] Step 3: Calculate information entropy and weights:

[0139] Construct an indicator ratio matrix: ;

[0140] Calculate information entropy: ;

[0141] For example, with n=100 demand points, the total population density index is calculated as follows: .

[0142] Calculate redundancy and weights: ;

[0143] Based on the existing data, the weights of each indicator are calculated as follows: .

[0144] Step 4: Calculate the overall score and classification:

[0145] Calculate the overall score: ;

[0146] For example, a certain street =0.1667×0.462+0.1794×0.556+...=0.688.

[0147] Sort by overall score in descending order, and categorize by quintiles: Very High Demand ( ≥0.8, weight 5), high demand (0.6≤ <0.8, weight 4), medium demand (0.4≤ <0.6, weight 3), low demand (0.2≤ <0.4, weight 2), very low demand ( <0.2, weight 1).

[0148] For example, a certain street =0.688, belonging to the high demand level, final weight =4.

[0149] Step 5: Result Verification and Application

[0150] The calculated demand point weights are used as input parameters for the subsequent site selection optimization model. The model's rationality is verified based on the overall evaluation results, and the weight data is then added to the demand database.

[0151] (II) Joint Optimization Method for Addressing and Scheduling in Deterministic Networks:

[0152] This example constructs a mixed-integer linear programming model. The specific steps are as follows:

[0153] Step 1: System Parameter Definition

[0154] Define the set of candidate take-off and landing points (Coordinates, Cost), Set of Demand Points (Coordinates, Weight), Airport Set (Type of aircraft equipped) Performance parameters ( , ), Hospital Collection (Coordinates), No-fly zone set (Center, Radius), System Constraint Parameters (Budget) Rescue time threshold Safe distance Ground speed ).

[0155] Step 2: Distance and Time Calculation

[0156] The flight distance takes into account detours around no-fly zones, with the following detour distance increments: ;

[0157] Flight time: (minute);

[0158] For example, if the actual distance is 18.77km and the speed of the medical aircraft is 210km / h, then... ≈5.36 minutes.

[0159] Step 3: Establish an optimization model:

[0160] Define decision variables: (Landing and takeoff site selection) (Demand Allocation) and (Patient's choice of transportation mode) and (Hospital selection) (Aircraft allocation) (auxiliary variables) and (Rescue time)

[0161] Objective function: ;

[0162] Key constraints:

[0163] Cost constraints: ;

[0164] Uniqueness of transportation mode: ;

[0165] Demand allocation constraints: , ;

[0166] Hospital allocation constraints: , ;

[0167] Aircraft allocation constraints: ;

[0168] Rescue time constraints: , , , , , , ;

[0169] Demand point weight time threshold constraint: ;

[0170] No-fly zone safety distance constraints: ;

[0171] Flight range constraints: ;

[0172] Objective function constraints: .

[0173] (III) Large-scale problem-solving methods based on the improved ALNS algorithm:

[0174] Step 1: Initial Solution Generation

[0175] (1) K-means clustering to determine take-off and landing points:

[0176] Number of clusters: ;

[0177] Randomly initialize k cluster centers and perform Lloyd iteration: calculate the distance from the required point to the cluster center and assign it to the nearest center; recalculate the cluster centers as the mean coordinates of the required points within each cluster; calculate the sum of squares (WCSS) within each cluster. Termination condition: center movement. The algorithm can run at km or iterate 300 times, or keep the allocation unchanged. Repeat this process 10 times, selecting the WCSS with the smallest result. Map the virtual center to the nearest candidate take-off and landing point to obtain the initial layout. .

[0178] (2) Greedy strategy demand allocation:

[0179] For each demand point: Calculate the direct transport time. If the journey takes ≤15 minutes, prioritize direct transport; otherwise, assess air transfer and calculate all possible configurations. Choose the configuration with the shortest time; compare the two modes and choose the better solution.

[0180] Initial solution objective value: ;

[0181] Step 2: Destroy operator design;

[0182] (1) Random destruction operator :

[0183] Damage ratio: ;

[0184] Quantity of damage: ;

[0185] Random selection The removal of a take-off and landing point disrupts the allocation of related demands.

[0186] (2) Bottleneck-oriented destruction operator :

[0187] Statistics on the number of take-off and landing point service demands and average arrival time Sorted by performance metrics, those causing disruption are ranked higher. The time required to decompose the demand points is... Determine the bottleneck type: If For "ground"; if If it is "flying," then it is "mixed." Count the number of bottlenecks of each type, determine the main types, and classify the damage according to severity.

[0188] Step 3: Repair operator design:

[0189] (1) Greedy repair operator R1:

[0190] For unallocated demand, evaluate direct transport and transshipment one by one in descending order of weight, and calculate the impact on the objective function. Choose the scheme that minimizes ΔT.

[0191] (2) k-regret repair operator R2:

[0192] Generate a set of all feasible solutions for each requirement. Sort by time and calculate the k-regret value: ;

[0193] For example, if k=3, and the program times are {13.2, 15.8, 16.5, ...}, then... (15.8-13.2)+(16.5-13.2)+...=9.5 minutes.

[0194] according to Sort in descending order, prioritizing requests with high regret values. If landing and takeoff points are not yet enabled, select and enable them near the request.

[0195] Step 4: Operator weight update:

[0196] Initialize weight w i =1. Roulette wheel selection, probability. .

[0197] Updated every 50 iterations: = ;

[0198] In the formula, =0.1; Number of successful attempts; This represents the number of uses. For example, if it is used 265 times and succeeded 72 times, then update w = 1.0 × [1 + 0.1 × (72 / 265)] = 1.027.

[0199] Step 5: Variable Neighborhood Descent Search

[0200] VND is executed every 100 ALNS iterations. Four types of neighborhoods are defined: (Adjustment of Takeoff and Landing Points), (Optimization of Hospitals), (Optimization of Aircraft Configuration), (Mode Switching).

[0201] VND Process: Initialization ; In Search for improvement; if found, update the solution and reset ; Otherwise ; When terminate.

[0202] Step Six, Simulated Annealing Acceptance:

[0203] Initial temperature , Cooling coefficient .

[0204] Acceptance probability: P = {1, if Δf < 0; , if Δf ≥ 0};

[0205] For example, Δf = 0.7, = 85, then P ≈ 0.992. Accept if the random number r < P. Temperature update ← × .

[0206] Step Seven, Elite Pool Management:

[0207] Capacity = 10. Calculate the solution similarity: ;

[0208] Adaptive diversity threshold: ;

[0209] If then add. When the capacity is full, remove the inferior solution with the target value , or remove the worse one in the most similar solution pair.

[0210] Step Eight, Algorithm Termination and Output:

[0211] Termination condition: iter ≥ 5000 or ≥ 500;

[0212] Select the optimal solution from the elite pool: .

[0213] Based on the optimal solution, output the takeoff and landing point layout, demand rescue plan, algorithm performance statistics, and operator usage statistics. Finally, supplement the optimized plan data to the plan database.

[0214] Through the above examples, a complete methodology for the selection and scheduling optimization of temporary take-off and landing sites for medical rescue aircraft has been constructed.

[0215] In summary, this embodiment provides an optimized air-ground collaborative medical rescue method that optimizes the selection of temporary take-off and landing points, the method of medical transport to demand points (direct ground transfer / air transfer), the aircraft flight route (airport → take-off and landing point → hospital), and the allocation relationship between demand points and take-off and landing points through joint decision-making.

[0216] This embodiment uses a parallel time model max{ground transport time, aircraft flight time} + transfer time to characterize the response time calculation method for simultaneous patient ground transport and aircraft flight.

[0217] This embodiment achieves a linearized processing method for detouring around no-fly zones by setting a minimum safe distance constraint between temporary take-off and landing points and the center of the no-fly zone;

[0218] This embodiment is a method for objectively weighting and evaluating potential medical rescue needs by constructing a multi-dimensional indicator system (population density, elderly population, takeoff safety factor, obstacle density, medical accessibility, GDP, and transportation convenience) based on the entropy weight method.

[0219] This embodiment identifies ground / flight / hybrid bottleneck types through a bottleneck-oriented destruction operator, prioritizes the demand with the largest regret value by combining a k-regret repair operator, and periodically embeds an enhanced adaptive large neighborhood search algorithm for local improvement using a variable neighborhood descent method.

[0220] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search, characterized in that, include: Based on multi-dimensional basic data of urban areas, obtain the weighted evaluation results of each medical rescue demand point; Based on the weight evaluation results, a hybrid integer programming model for site selection and demand allocation is constructed. Based on the aforementioned mixed integer programming model, an initial feasible solution is obtained through K-means clustering and a greedy heuristic algorithm. Based on the initial feasible solution, the optimal solution is obtained through iterative operations of the destruction operator and the repair operator, as well as local search using the variable neighborhood descent method; Based on the optimal solution, a temporary take-off and landing point layout scheme and an emergency rescue dispatch scheme are obtained.

2. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 1, characterized in that, Based on multi-dimensional basic data of the urban area, the weighted evaluation results of each medical rescue demand point are obtained, including: Based on data on population density, number of elderly people, takeoff safety factor, obstacle density, medical accessibility distance, GDP level, and transportation convenience, a standardized data matrix is ​​obtained after positive index processing. Based on the standardized data matrix, the information entropy value and redundancy of each indicator are calculated using the entropy weight method to obtain the objective weight value of each indicator. Based on the objective weight values ​​of each indicator and the standardized data matrix, a comprehensive score value for each requirement point is obtained by linear weighted summation. The weighted evaluation result is obtained based on the comprehensive score.

3. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 2, characterized in that, Based on the comprehensive score, the weighted evaluation result is obtained as follows: The comprehensive score values ​​are sorted in descending order; The overall score after ranking is divided into preset levels and assigned corresponding weight values. The weight evaluation results are obtained and the corresponding weight values ​​are used as quantitative parameters of the importance of demand points in the subsequent site selection optimization model, and as the basis for the grid-based division of demand points in the corresponding city.

4. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 1, characterized in that, Based on the weight evaluation results, the site selection-demand allocation hybrid integer programming model is constructed as follows: Based on the candidate take-off and landing point set, demand point set, hospital set, aircraft performance parameters, and no-fly zone data, a decision variable system is obtained; Based on the aforementioned decision variable system and the rescue response time calculation rules for each demand point, the objective function is obtained; Based on the evaluation results of the demand points, the construction cost of take-off and landing points, the allocation relationship of demand points, the aircraft scheduling rules, the safety distance of no-fly zones and the flight range of aircraft, obtain the set of constraints. Based on the objective function and the set of constraints, the location-demand allocation hybrid integer programming model is obtained.

5. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 4, characterized in that, The rules for calculating the rescue response time at each demand point include: For demand points that utilize direct ground transport, the rescue response time is the ground transport time from the demand point to the target point. For demand points that require air transport, the rescue response time is determined by the greater of the ground time from the demand point to the temporary take-off and landing point and the flight time of the aircraft from the airport to the temporary take-off and landing point, plus the flight time from the temporary take-off and landing point to the target point.

6. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 1, characterized in that, Based on the aforementioned mixed-integer programming model, the initial feasible solutions are obtained through K-means clustering and a greedy heuristic algorithm, including: Based on the aforementioned mixed integer programming model, K-means clustering is used to determine the take-off and landing points; Based on the take-off and landing points, a greedy heuristic algorithm is used to determine the path with the shortest rescue time for each demand point, thereby obtaining the initial feasible solution.

7. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 1, characterized in that, Based on the initial feasible solution, the optimal solution is obtained through iterative operations of the destruction and repair operators and local search using the variable neighborhood descent method, including: The initial feasible solution is destroyed using the destruction operator to obtain a destroyed solution, wherein the destruction operator includes a random destruction operator and a bottleneck-oriented destruction operator; Based on the destructive solution, the unallocated demand points are reallocated using the repair operator to obtain a repair solution, wherein the repair operator includes a greedy repair operator and a k-regret repair operator; Based on the repaired solution, the optimal solution is obtained by combining the local search of the variable neighborhood descent method.

8. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 7, characterized in that, The bottleneck-oriented destruction operator is used to destroy the initial feasible solution to obtain a destroyed solution, including: Based on the number of service demands and average arrival time of each take-off and landing point in the initial feasible solution, obtain the performance ranking results of the take-off and landing points; Based on the ground transportation time, flight time, and total rescue time of each demand point in the initial feasible solution, obtain the bottleneck type identification result; Based on the performance ranking results and the bottleneck type identification results, take-off and landing points or bottleneck demand points with poor performance are removed first to obtain the destructive solution.

9. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 7, characterized in that, Based on the destructive solution, the unallocated demand points are reallocated using the k-regret repair operator to obtain the repair solution, including: Based on the rescue time from each unassigned demand point to each take-off and landing point and hospital, obtain the regret values ​​of k optimal solutions; Based on the sorting results of the regret values ​​from largest to smallest, the rescue plan is allocated to the demand point with the largest regret value first, and the repair solution is obtained.

10. The method for optimizing the location and scheduling of low-altitude medical rescue aircraft based on adaptive large neighborhood search according to claim 7, characterized in that, Obtaining the optimal solution also includes: Based on the historical success count and usage count of each destruction operator and repair operator, obtain the operator weight update value; Based on the updated operator weight values, the destruction and repair operators to be used in the next iteration cycle are obtained through a roulette wheel selection mechanism.

Images