A Method for Selecting the Location of a Drone Nest for Rapid Off-site Evidence Collection in Traffic

Abstract

The present invention discloses a nest site selection method for non - on - site rapid evidence collection of traffic for unmanned aerial vehicles, belonging to the technical field of unmanned aerial vehicle nest site selection. The method is as follows: First, find accident - prone intersections, obtain and process traffic data of intersections within the selected area, use the entropy weight method to assign weights to each index in the data, use the TOPSIS method to calculate the comprehensive scores of each intersection, establish a risk threshold, establish accident - prone intersections as risk points and assign weights, for the accident risk points obtained in the above steps, establish a plane rectangular coordinate system, map the risk points to the coordinate system, and obtain their position coordinates; Then, establish a non - on - site evidence collection unmanned aerial vehicle nest site selection model for traffic accidents, solve the unmanned aerial vehicle nest site selection model based on the NSGA - II algorithm, and obtain the optimal nest site selection scheme.

Description

Technical Field

[0001] The present invention belongs to the technical field of UAV nest site selection, and particularly relates to a nest site selection method for non-site rapid evidence collection in traffic oriented to UAVs. Background Art

[0002] With the rapid increase in the number of motor vehicles in China, the incidence of traffic accidents has also shown an upward trend. At present, traffic accident evidence collection mainly relies on traffic police arriving at the scene, taking pictures of the details of the accident scene, and then making judgments in combination with the video of road network cameras. When the accident rate is high, due to the time required for traffic police to arrive at the scene and the limited police resources, it often leads to low efficiency in accident handling, increased traffic congestion, and even an increased risk of secondary accidents.

[0003] In recent years, some cities have tried to introduce UAVs to assist in evidence collection. By quickly locating the accident scene from an aerial perspective and collecting image evidence, it has alleviated the ground police force pressure to a certain extent. However, most of the existing UAV evidence collection modes adopt the manual remote control operation mode. Although it can replace some on-site investigation work, in essence, it still requires special personnel to operate, cannot break through the human resource bottleneck, and has certain limitations.

[0004] A UAV nest is an automated aviation infrastructure designed specifically for UAVs. Inside the nest, there are equipped with automatic charging stations, data transmission devices, etc. UAVs can achieve automatic takeoff and landing, automatic charging, and automatic maintenance inside the nest. The nest can collect, store, and analyze the data transmitted by UAVs to provide support for decision-making. In case of emergencies, the UAV nest can quickly deploy UAVs for aerial patrol or rescue. It is widely used in fields such as power line inspection, traffic monitoring, and security patrol, which can replace manual operations, improve work efficiency, and reduce labor costs and safety risks.

[0005] Currently, the site selection of UAV nests is mostly used for power supply station line inspection, highway road administration maintenance, etc., and there are few site selections for non-site rapid evidence collection scenarios in traffic. In complex traffic scenarios, the site selection of the nest needs to comprehensively consider various factors such as traffic flow and accident-prone areas to ensure that the UAV can respond quickly and accurately collect evidence. However, the existing site selection methods mostly use subjective methods, lacking a scientific evaluation system and quantitative indicators, and it is difficult to meet the actual needs of traffic evidence collection. Summary of the Invention

[0006] In view of the above problems existing in the prior art, the present invention proposes a nest site selection method for non-site rapid evidence collection in traffic oriented to UAVs, with reasonable design, which solves the deficiencies of the prior art and has good effects.

[0007] A nest site selection method for non-site rapid evidence collection in traffic oriented to UAVs includes the following steps:

[0008] S1: Search for intersections with high accident rates;

[0009] S2: Establish a siting model for non - on - site evidence - collection drones at traffic accident scenes;

[0010] S3: Solve the siting model for drones based on the NSGA - II algorithm to obtain the optimal drone siting plan.

[0011] Further, the S1 includes the following sub - steps:

[0012] S1.1: Obtain and process traffic data of intersections within the selected area;

[0013] S1.2: Use the entropy weight method to assign weights to each index in the data;

[0014] S1.3: Use the TOPSIS method to calculate the comprehensive scores of each intersection;

[0015] S1.4: Establish a risk threshold, identify intersections with high accident rates as risk points and assign weights;

[0016] S1.5: For the accident risk points obtained in the above steps, establish a plane rectangular coordinate system, map the risk points into the coordinate system, and obtain their position coordinates .

[0017] Further, the S1.1 is specifically: For all intersections within the selected area, collect and organize the following data:

[0018] The intersection set is defined as:

[0019] ;

[0020] Among them, and represent negative indicators, which are the number of lanes and the sight - distance score respectively, represents positive indicators, which are the annual accident number, daily average traffic flow, average vehicle speed, hard - braking frequency, and pedestrian flow respectively;

[0021] The original data matrix is defined as:

[0022] ;

[0023] Among them, represents the original value of the th intersection at the th index;

[0024] Standardize the collected data:

[0025] For positive indicators:

[0026] , ;

[0027] For negative indicators:

[0028] , ;

[0029] Among them, is the value of the th intersection after standardization for the th indicator, is the maximum value in the original values of the th indicator, is the minimum value in the original values of the th indicator;

[0030] The specific content of S1.2 is as follows:

[0031] Calculate the indicator ratio, and the expression is:

[0032] ;

[0033] Among them, is the ratio of the th intersection corresponding to the th indicator, is the value of the th intersection after standardization for the th indicator, ;

[0034] Calculate the information entropy, and the expression is:

[0035] ;

[0036] Among them, is the information entropy of the th indicator, is the ratio of the th intersection corresponding to the th indicator;

[0037] When , define ;

[0038] Calculate the coefficient of variation, and the expression is:

[0039] ;

[0040] Among them, is the coefficient of variation of the th indicator;

[0041] Determine the weights of each indicator, and the expression is:

[0042] ;

[0043] Among them, is the weight of the th index, is the coefficient of variation of the th index.

[0044] Furthermore, the S1.3 is specifically: construct a weighted matrix, and the element in the weighted matrix is expressed as:

[0045] ;

[0046] Define the ideal solution:

[0047] The expression of the positive ideal solution corresponding to the th index is:

[0048] ;

[0049] The expression of the negative ideal solution corresponding to the th index is:

[0050] ;

[0051] Calculate the distance:

[0052] The Euclidean distance from the th intersection to the positive ideal solution is expressed as:

[0053] ;

[0054] The Euclidean distance from the th intersection to the negative ideal solution is expressed as:

[0055] ;

[0056] Calculate the closeness, and the expression is:

[0057] ;

[0058] Among them, is the closeness corresponding to the th intersection;

[0059] Output the comprehensive score vector :

[0060] ;

[0061] The 1.4 specifically is as follows: For all intersections that already have their respective comprehensive scores, first determine the threshold setting, and the expression is:

[0062] ;

[0063] Among them, is the demand sensitivity, is the highest score, is the lowest score;

[0064] Using the threshold as the boundary, screen out the intersections with scores higher than the threshold as high-accident intersections, defined as risk points, and assign weights to each risk point , and the expression is:

[0065] ;

[0066] Among them, is the weight of the th risk point, is the score of the th risk point, is the number of risk points.

[0067] Furthermore, the S2 includes the following sub-steps:

[0068] S2.1: Define the model objective function. The goal of site selection is to maximize the coverage area of the drone nest and minimize the total construction cost of the drone nest. The expression is:

[0069] ;

[0070] ;

[0071] Among them, each risk point has coordinates , Q is the number of alternative points, representing the alternative locations of the drone nest. Each alternative point has coordinates ; Take the risk points as the alternative points for drone nest site selection at the same time to maximize the coverage area, is the fixed cost of the drone nest, is the incremental cost of the drone nest with a capacity of , is the incremental cost of the drone nest with a capacity of , is a 0-1 decision variable indicating whether to build a drone nest with a capacity of at the th alternative point, is a 0-1 variable indicating whether to build a drone nest with a capacity of at the th alternative point; is a variable indicating the Whether the \(i\)-th risk point is covered by the \(j\)-th alternative point's hangar, a 0-1 decision variable;

[0072] S2.2: Determine the coverage relationship matrix between risk points and alternative points:

[0073] Establish a \(r\times q\) order 0-1 matrix \(A\), defined as the coverage relationship matrix between risk points and alternative points, and calculate each element \(a_{ij}\) in the matrix according to the following formula: rq :

[0074] ;

[0075] where, is the maximum communication distance of the UAV, is the maximum flight distance within the UAV's endurance, and its value follows the following formula:

[0076] ;

[0077] where, is the longest endurance time of the UAV, is the time taken for the UAV to perform a single evidence collection task, is the flight speed of the UAV, is the average flight altitude of the UAV;

[0078] S2.3: Define the key constraints of the model:

[0079] Coverage feasibility, the expression is:

[0080] ;

[0081] Ensure that its coverage meets the coverage distance limit. Only when the \(j\)-th alternative point builds a hangar and its distance from the \(i\)-th risk point is within the coverage radius, is it considered that the \(i\)-th risk point is covered by the \(j\)-th alternative point's hangar;

[0082]

[0083] Unique coverage, the expression is:

[0083] ;

[0084] Ensure that each risk point is covered by at least one hangar;

[0085] Capacity limit, the expression is:

[0086] ;

[0087] Ensure that the hangar at can meet the range​ Demand of points

[0088] Mutual exclusion constraint, the expression is:

[0089] ;

[0090] Ensure that only one nest is built at each selected site point.

[0091] Furthermore, the said S3 includes the following sub-steps:

[0092] S3.1: Calculate and screen out eligible risk intersections and obtain their coordinates, input relevant parameters according to the model, input relevant data and calculate the coverage radius matrix; Set the parameters of the NSGA-II algorithm, including the population size N, the maximum number of iterations T, the crossover probability P c , and the mutation probability P m ;

[0093] S3.2: The chromosome consists of two parts: the nest construction decision segment and the coverage relationship decision segment. Among them, the nest construction decision segment uses binary coding, with a length of 2|Q|. The first |Q| bits represent the decision segment X for whether to build a nest with a capacity of aq , and the last |Q| bits represent the decision segment X for whether to build a nest with a capacity of bq ; The coverage relationship decision segment uses binary matrix coding Z rq , indicating whether the th risk point is covered by the th alternative point, with a dimension of R×Q, and needs to meet the coverage feasibility constraint;

[0094] Randomly generate an initial population with a size of N, but allow the chromosome to temporarily violate the mutual exclusion constraint and the unique coverage constraint;

[0095] S3.3: For each individual in the population, define a two-objective fitness function:

[0096] ;

[0097] ;

[0098] Among them, is the fitness function of the first objective, is the fitness function of the second objective;

[0099] Introduce a multi-constraint violation penalty mechanism to comprehensively handle the following constraints:

[0100] ;

[0101] The penalty term is designed as:

[0102] ;

[0103] In the formula, , , are penalty coefficients, which are dynamically adjusted according to the constraint priority;

[0104] Attach the penalty term to the objective function to form a modified bi-objective fitness function:

[0105] ;

[0106] ;

[0107] Among them, is the fitness function of the first modified objective, is the fitness function of the first modified objective; after modification, it can ensure that individuals violating the constraints are gradually eliminated during the evolution process;

[0108] S3.4: Perform non-dominated sorting on all individuals in the population to divide the Pareto front rank of the individuals; calculate the crowding degree of the individuals within the same front layer;

[0109] S3.5: Perform genetic operations, which are divided into three parts: selection, crossover, and mutation;

[0110] For the selection operation, the tournament selection method is adopted. Randomly select K individuals from the population, compare their non-dominated ranks and crowding degrees, and preferentially select individuals with a lower front rank. If they are in the same rank, select individuals with a larger crowding degree to maintain diversity;

[0111] For the crossover operation, perform two-point crossover on the nest construction decision segment. Randomly select two crossover points and exchange the corresponding segments of the nest decision segment in the parental chromosomes. The coverage relationship segment is dynamically adjusted according to the crossover result to update the coverage relationship matrix of the offspring; ensure that:

[0112] ;

[0113] In the mutation operation, with a mutation probability randomly flip a certain bit in X aq or X aq . If the state of a certain nest changes after flipping, the corresponding element in the coverage relationship Z rq of this nest needs to be forced to be set to 0;

[0114] S3.6: Merge the parental population and the offspring population. The size of the merged population is 2N. Re-perform non-dominated sorting and crowding degree calculation on the merged population; retain the top N optimal individuals in the order of the front as the new population;

[0115] S3.7: If the maximum number of iterations T is reached, terminate the algorithm and output the Pareto front solution set; otherwise, return to S3.4;

[0116] The Pareto front solution set includes different site selection plan diagrams and their corresponding costs and coverage risk indices considering the penalty coefficient, and generates a coordinate scatter plot that can reflect the two index values; from the Pareto front solution set, according to specific decision-making requirements and preferences, if the decision maker needs to achieve the best coverage efficiency without considering costs, select the plan with the highest coverage risk index considering the penalty coefficient in the Pareto front solution; conversely, if the decision maker needs the plan with the lowest cost and can thus sacrifice some coverage effect, select the plan with the lowest cost in the Pareto front solution; if the decision maker needs to balance the two indices for comprehensive consideration, select other front solutions in the Pareto front solution set, make a decision by referring to the coordinate scatter plot and the site selection plan diagram, so as to select the optimal drone nest site selection plan.

[0117] The beneficial technical effects brought by the present invention:

[0118] The present invention proposes a method for site selection of drone nests for non-field rapid evidence collection in traffic. After obtaining relevant data of traffic intersections in the area and relevant parameters of drone nests, aiming at the regional characteristics, it can realize an accurate and efficient multi-objective optimization technology for site selection of drone nests, which can include two different capacities of drone nests, comprehensively consider costs and coverage effects, and finally can output multiple site selections biased towards different indices for decision makers to refer to, with good scientificity, reliability and robustness, and can be applied to non-field evidence collection management in traffic. Brief Description of the Drawings

[0119] Figure 1 It is a flow chart of a method for site selection of drone nests for non-field rapid evidence collection in traffic according to the present invention.

[0120] Figure 2 It is a schematic diagram of risk points in a plane rectangular coordinate system in Embodiment 1.

[0121] Figure 3 It is a schematic diagram of five feasible solutions in the Pareto front solution set in Embodiment 1.

[0122] Figure 4 It is a site selection diagram of Solution 1 (11.74, 3090059.64) in Embodiment 1.

[0123] Figure 5 It is a site selection diagram of Solution 2 (14.73, 4305056.66) in Embodiment 1.

[0124] Figure 6Location map for Solution 3 (13.74, 3210057.64) in Example 1.

[0125] Figure 7 Location map for Solution 4 (14.43, 4275056.96) in Example 1.

[0126] Figure 8 Location map for Solution 5 (15.26, 5040056.13) in Example 1.

[0127] Figure 9 Curve graph of the total cost versus the number of iterations in Example 1.

[0128] Figure 10 Curve of the coverage risk index after considering the comprehensive penalty coefficient versus the number of iterations in Example 1. Detailed implementation manners

[0129] The following further describes the detailed implementation manners of the present invention in conjunction with specific embodiments:

[0130] A nest location method for non - on - site rapid evidence collection of traffic for unmanned aerial vehicles, comprising the following steps:

[0131] S1: Search for accident - prone intersections, including the following sub - steps:

[0132] S1.1: Obtain and process intersection traffic data within the selected area;

[0133] For all intersections within the selected area, collect and collate the following data:

[0134] The intersection set is defined as:

[0135] ;

[0136] Among them, and represent negative indicators, which are the number of lanes and the sight - distance score respectively, represents positive indicators, which are the number of annual accidents, daily average traffic flow, average vehicle speed, hard - braking frequency, and pedestrian flow respectively;

[0137] The original data matrix is defined as:

[0138] ;

[0139] Among them, represents the original value of the th intersection at the th indicator;

[0140] Standardize the collected data. The goal of standardization is to map the data to the interval [0, 1] to avoid the influence of differences in units or magnitudes of different indicators on the fairness of subsequent weight allocation. The standardization rules are as follows:

[0141] For positive indicators:

[0142] , ;

[0143] For negative indicators:

[0144] , ;

[0145] Among them, is the value of the th intersection for the th indicator, is the maximum value among the original values of the th indicator, is the minimum value among the original values of the th indicator.

[0146] S1.2: Use the entropy weight method to allocate weights to each indicator in the data;

[0147] The entropy weight method is an objective weight assignment method based on the principle of information theory. It determines the weights by calculating the information entropy of each indicator and is often used in multi-index comprehensive evaluation and decision-making analysis. In the present invention, the weight allocation of the entropy weight method to the data can be carried out in the following four steps:

[0148] ① Calculate the indicator proportion, and the expression is:

[0149] ;

[0150] Among them, is the proportion of the th intersection corresponding to the th indicator, is the value of the th intersection after standardization for the th indicator, ;

[0151] ② Calculate the information entropy, and the expression is:

[0152] ;

[0153] Among them, is the information entropy of the th indicator, is the th intersection corresponding to the Proportion of each index;

[0154] When it is defined as ;

[0155] ③ Calculate the coefficient of variation, and the expression is:

[0156] ;

[0157] Among them, is the coefficient of variation of the th index;

[0158] ④ Determine the weight of each index, and the expression is:

[0159] ;

[0160] Among them, is the weight of the th index, is the coefficient of variation of the th index.

[0161] S1.3: Use the TOPSIS method to calculate the comprehensive score of each intersection;

[0162] The TOPSIS method is a multi-criteria decision analysis method. By constructing the positive ideal solution (optimal solution) and negative ideal solution (worst solution) of the evaluation problem, calculating the similarity between each candidate solution and the ideal solution and negative ideal solution, evaluating and ranking the advantages and disadvantages of each solution, and finally selecting the solution closest to the ideal solution as the optimal solution. In the present invention, the TOPSIS method scores each intersection according to the following four steps:

[0163] Construct a weighted matrix, and the element in the weighted matrix has the following expression:

[0164] ;

[0165] Define the ideal solution:

[0166] The expression for the positive ideal solution corresponding to the th index is:

[0167] ;

[0168] The expression for the negative ideal solution corresponding to the th index is:

[0169] ;

[0170] Calculate the distance:

[0171] The Euclidean distance from the th intersection to the positive ideal solution

[0172] ;

[0173] The Euclidean distance from the th intersection to the negative ideal solution

[0174] ;

[0175] Calculate the closeness, and the expression is:

[0176] ;

[0177] where is the closeness corresponding to the th intersection;

[0178] Output the comprehensive scoring vector :

[0179] .

[0180] S1.4: Establish the risk threshold, identify the intersections with high accident rates as risk points and assign weights, specifically:

[0181] For all intersections with their respective comprehensive scores, first determine the threshold setting, and the expression is:

[0182] ;

[0183] where is the demand sensitivity, which can be set as α = 0.6 and adjusted according to the demand, is the highest score, is the lowest score;

[0184] Select the intersections with scores higher than the threshold as the intersections with high accident rates, define them as the high-risk points of traffic accidents, and assign weights to each risk point , and the expression is:

[0185] ;

[0186] where is the weight of the th risk point, is the score of the th risk point, is the number of risk points.

[0187] S1.5: For the accident risk points obtained in the above steps, establish a plane rectangular coordinate system, map the risk points into the coordinate system, and obtain their position coordinates. 。

[0188] S2: Establish a non - on - site evidence - collection UAV nest site - selection model for traffic accidents, including the following sub - steps:

[0189] S2.1: Define the model objective function;

[0190] This model is a multi - objective programming model. The goal of site - selection is to maximize the coverage area of the nest and minimize the total construction cost of the nest. The expression is:

[0191] ;

[0192] ;

[0193] Among them, each risk point has coordinates , Q is the number of alternative points, representing the alternative positions of the UAV nest. Each alternative point has coordinates ; The risk points are simultaneously used as alternative points for nest site - selection to maximize the coverage area. is the fixed cost of the nest, is the incremental cost of the nest with capacity , is the incremental cost of the nest with capacity , is the 0 - 1 decision variable indicating whether to build a nest with capacity at the th alternative point, is the 0 - 1 variable indicating whether to build a nest with capacity at the th alternative point; is the 0 - 1 decision variable indicating whether the th risk point is covered by the nest at the th alternative point;

[0194] S2.2: Determine the coverage relationship matrix between risk points and alternative points;

[0195] Establish an r×q - order 0 - 1 matrix A, define it as the coverage relationship matrix between risk points and alternative points, and calculate each element a in the matrix according to the following formula rq :

[0196] ;

[0197] Among them, is the maximum communication distance of the UAV, The maximum flight distance within the endurance of the UAV, and its value follows the following formula:

[0198] ;

[0199] where, is the longest endurance time of the UAV, is the time taken for the UAV to perform a single evidence collection task, is the flight speed of the UAV, is the average flight altitude of the UAV;

[0200] S2.3: Define the key constraints of the model:

[0201] This model comprehensively considers the endurance ability of the UAV and the characteristics of other related problems, and sets the following constraints to ensure that the model meets the requirements of the actual application scenario.

[0202] ① Coverage feasibility, the expression is:

[0203] ;

[0204] Ensure that its coverage meets the coverage distance limit. Only when the th alternative point builds a hangar and its distance from the th risk point is within the coverage radius, is it considered that the th risk point is covered by the hangar at the th alternative point;

[0205] ② Unique coverage, the expression is:

[0206] ;

[0207] Ensure that each risk point is covered by at least one hangar;

[0208] ③ Capacity limit, the expression is:

[0209] ;

[0210] Ensure that the hangar at can meet the requirements of the points within the range of ;

[0211] ④ Mutual exclusion constraint, the expression is:

[0212] ;

[0213] Ensure that only one hangar is built at each selected site.

[0214] S3: Solve the UAV hangar location model based on the NSGA-II algorithm to obtain the optimal hangar location plan, including the following sub-steps:

[0215] S3.1: Parameter initialization and data preprocessing;

[0216] Calculate and screen out eligible risk intersections and obtain their coordinates. According to the relevant parameters of the model, input relevant data and calculate the coverage radius matrix; set the parameters of the NSGA-II algorithm, including the population size N, the maximum number of iterations T, the crossover probability , and the mutation probability ;

[0217] S3.2: Chromosome encoding and population initialization;

[0218] The chromosome consists of two parts: the decision segment for airship hangar construction and the coverage relationship segment. Among them, the decision segment for airship hangar construction uses binary encoding, with a length of 2|Q|. The first |Q| bits represent the decision segment X of whether to build an airship hangar with a capacity of aq , and the last |Q| bits represent the decision segment X of whether to build an airship hangar with a capacity of bq ; the coverage relationship segment uses binary matrix encoding Z rq , indicating whether the th risk point is covered by the th alternative point, with a dimension of R×Q, and it needs to satisfy the coverage feasibility constraint;

[0219] Randomly generate the initial population, with a size of N, but allow the chromosome to temporarily violate the mutual exclusion constraint and the unique coverage constraint;

[0220] S3.3: Fitness function and constraint handling;

[0221] For each individual in the population, define a two-objective fitness function:

[0222] ;

[0223] ;

[0224] Among them, is the fitness function of the first objective, is the fitness function of the second objective;

[0225] Introduce a multi-constraint violation penalty mechanism to comprehensively handle the following constraints:

[0226] ;

[0227] The penalty term is designed as:

[0228] ;

[0229] In the formula, , , is the penalty coefficient, which is dynamically adjusted according to the constraint priority;

[0230] Attach the penalty term to the objective function to form the modified bi-objective fitness function:

[0231] ;

[0232] ;

[0233] Among them, is the fitness function of the first modified objective, is the fitness function of the first modified objective; after modification, it can ensure that individuals violating the constraints are gradually eliminated during the evolution process;

[0234] S3.4: Non-dominated sorting and crowding degree calculation;

[0235] Perform non-dominated sorting on all individuals in the population to divide the Pareto front rank of the individuals; calculate the crowding degree of individuals within the same front layer to ensure population diversity;

[0236] S3.5: Perform genetic operations, and the genetic operations are divided into three parts: selection, crossover, and mutation;

[0237] The selection operation adopts the tournament selection method, randomly select K individuals from the population, compare their non-dominated ranks and crowding degrees, and preferentially select individuals with lower front ranks. If they are in the same rank, select individuals with larger crowding degrees to maintain diversity;

[0238] For the crossover operation, perform two-point crossover on the decision segment of the nest construction. Randomly select two crossover points and exchange the corresponding segments of the nest decision segment in the parental chromosomes. The coverage relationship segment is dynamically adjusted according to the crossover result to update the coverage relationship matrix of the offspring; ensure that:

[0239] ;

[0240] In the mutation operation, with a mutation probability randomly flip a certain bit in X aq or X aq . If the state of a certain nest changes after flipping, the corresponding element in the coverage relationship Z rq of this nest needs to be forced to be 0;

[0241] S3.6: Population update;

[0242] Merge the parental population and the offspring population. After merging, the population size is 2N. Re - perform non - dominated sorting and crowding degree calculation on the merged population; retain the top N optimal individuals in the order of fronts as the new population (if the total number of individuals in a certain front exceeds the remaining capacity, select them from high to low according to the crowding degree);

[0243] S3.7: Termination determination and Pareto front solution set output;

[0244] If the maximum number of iterations T is reached, terminate the algorithm and output the Pareto front solution set; otherwise, return to S3.4;

[0245] The Pareto front solution set contains different site - selection plan graphs and their corresponding costs and coverage risk indices considering the penalty coefficient, and generates a coordinate scatter plot that can reflect the two - index values; from the Pareto front solution set, according to specific decision - making requirements and preferences, if the decision - maker needs to achieve the best coverage efficiency without considering costs, select the plan with the highest coverage risk index considering the penalty coefficient in the Pareto front solution; conversely, if the decision - maker needs the plan with the lowest cost and can thus sacrifice some coverage effect, select the plan with the lowest cost in the Pareto front solution; if the decision - maker needs to balance the two indices for comprehensive consideration, select other front solutions in the Pareto front solution set, and can make decisions by referring to the coordinate scatter plot and the site - selection plan graph, so as to select the optimal nest site - selection plan.

[0246] Example 1:

[0247] In this example, a region with a size of 3000m * 3000m in a certain city is obtained, and all the required indicators of all intersections in it are obtained. After standardization, weight assignment by the entropy - weight method and score assignment by the TOPSIS method, the demand sensitivity is set to 0.6. After comprehensive calculation, 146 intersections exceeding the threshold are screened out, and their corresponding weights are assigned according to probability . Set them as risk points and draw them in the plane rectangular coordinate system as Figure 2 shown.

[0248] Set the fixed cost of the drone nest = 80000 yuan, the incremental cost = 10000 yuan, the incremental cost = 25000 yuan; the capacities of the two types of nests are respectively = 1, = 2; the maximum communication distance of the drone = 3000m, the maximum endurance time = 30min, the time taken for a single evidence - collection mission = 8min, the flight speed = 200 m / min, flight altitude = 200 m.

[0249] Under the Python 3.12 environment, the NSGA-II algorithm adopted by this model is programmed based on the Pycharm platform, relevant parameters are input, and the penalty coefficient is set = = = 1, population size N = 500, maximum number of iterations T = 600, crossover probability = 0.75, mutation probability = 0.05.

[0250] Running the algorithm finally obtains the Pareto front solution set as shown in Figure 3 shown. It can be seen that after removing duplicates, it has five feasible solutions, and the corresponding total cost and the coverage risk index after comprehensive penalty coefficient have been marked at each solution. The site selection map corresponding to each solution is as shown in Figures 4 - 8 shown. These site selection maps contain the construction reference positions of two types of hangars and the positions of all risk points, and the cost and coverage risk index corresponding to this solution are also marked. At the same time, the change curve of the total cost with the number of iterations is output as shown in Figure 9 shown; the change curve of the coverage risk index after comprehensive penalty coefficient with the number of iterations is as shown in Figure 10 shown.

[0251] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for selecting the location of a drone nest for non - on - site rapid evidence collection in traffic, characterized in that, It includes the following steps: S1: Search for accident-prone intersections; S2: Establish a non-site evidence collection UAV nest site selection model for traffic accidents; S3: Solve the UAV nest site selection model based on the NSGA-II algorithm to obtain the optimal nest site selection plan; The S2 includes the following sub-steps: S2.1: Define the model objective function. The goal of site selection is to maximize the coverage range of the nest and minimize the total cost of nest construction. The expression is: ； ； Among them, each risk point has coordinates , Q is the number of alternative points, representing the alternative locations of the UAV nests, and each alternative point has coordinates ; The risk points are simultaneously used as alternative points for nest site selection to maximize the coverage area, is the fixed cost of the nest, is the incremental cost of the nest with a capacity of , is the incremental cost of the nest with a capacity of , represents the 0-1 decision variable indicating whether to build a nest with a capacity of at the th alternative point, represents the 0-1 variable indicating whether to build a nest with a capacity of at the th alternative point; represents the 0-1 decision variable indicating whether the th risk point is covered by the nest at the th alternative point; S2.2: Determine the coverage relationship matrix between risk points and alternative points; Construct an \(r\times q\) 0-1 matrix \(A\), which is defined as the coverage relationship matrix between risk points and alternative points, and calculate each element \(a\) in the matrix according to the following formula rq : ； Among them, is the maximum communication distance of the drone, is the maximum flight distance within the endurance of the drone, and its value follows the following formula: ； Wherein, is the longest endurance time of the UAV, is the time taken for the UAV to perform a single evidence collection mission, is the flight speed of the UAV, is the average flight altitude of the UAV; S2.3: Define the key constraints of the model: Coverage feasibility, the expression is: ； Ensure that its coverage meets the coverage distance limit. Only when the th alternative point builds a nest and its distance from the th risk point is within the coverage radius, is the th risk point considered to be covered by the th alternative point nest; Unique coverage, the expression is: ； Ensure that each risk point is covered by at least one nest; Capacity limit, the expression is: ； Guarantee The hangar at can meet the requirements at points within the range; Mutual exclusion constraint, the expression is: ； Ensure that only one nest is built at each selected site.

2. The nest site selection method for non - on - site rapid evidence collection of traffic for unmanned aerial vehicles according to claim 1, wherein, The S1 includes the following sub-steps: S1.1: Obtain and process the intersection traffic data within the selected area; S1.2: Use the entropy weight method to assign weights to each index in the data; S1.3: Use the TOPSIS method to calculate the comprehensive scores of each intersection; S1.4: Establish a risk threshold, identify accident-prone intersections as risk points and assign weights; S1.5: For the accident risk points obtained in the above steps, establish a rectangular coordinate system, map the risk points to the coordinate system, and obtain their position coordinates .

3. The method for selecting a nest location for non-site rapid evidence collection of traffic for unmanned aerial vehicles according to claim 2, wherein, The specific content of S1.1 is: For all intersections within the selected area, collect and organize the following data: The intersection set is defined as: ； Among them, and represent negative indicators, namely the number of lanes and the sight distance score respectively, represent positive indicators, namely the annual accident number, the daily average traffic flow, the average vehicle speed, the hard braking frequency, and the pedestrian flow respectively; The original data matrix is defined as: ； Among them, represents the -th intersection's original value for the -th indicator; Standardize the collected data: For positive indicators: ， ； For negative indicators: ， ； Among them, is the value of the th intersection after standardization processing for the th indicator, is the maximum value among the original values of the th indicator, is the minimum value among the original values of the th indicator; The specific content of S1.2 is: Calculate the index proportion, the expression is: ； Among them, is the th proportion of the th index corresponding to the th intersection after standardization, is the value of the th index of the ; Calculate the information entropy, the expression is: ； Among them, is the information entropy of the th index, is the proportion of the th index corresponding to the th intersection; When is defined as ; Calculate the difference coefficient, the expression is: ； Among them, is the coefficient of variation of the th index; Determine the weights of each index, the expression is: ； Among them, is the weight of the th index, is the coefficient of variation of the th index.

4. A nest site selection method for non - on - site rapid evidence collection of traffic for unmanned aerial vehicles according to claim 3, characterized in that, The specific content of S1.3 is as follows: construct a weighted matrix, and the elements in the weighted matrix The expression is: ； Define the ideal solution: The positive ideal solution expression corresponding to the item index is as follows: ； The negative ideal solution expression corresponding to the index item is as follows: ； Calculate the distance: The Euclidean distance from the first intersection to the positive ideal solution is expressed as: ； The Euclidean distance from the ith intersection to the negative ideal solution is expressed as: ； Calculate the closeness, the expression is: ； Among them, is the proximity corresponding to the th intersection; Output comprehensive scoring vector : ； The specific content of 1.4 is: For all intersections that already have their respective comprehensive scores, first determine the threshold setting, the expression is: ； Among them, is the demand sensitivity, is the highest score, is the lowest score; Intersections with scores higher than the threshold are screened with the threshold as the boundary and defined as high-incidence accident intersections, i.e., risk points, and weights are assigned to each risk point , and the expression is: ； Among them, is the weight of the th risk point, is the score of the th risk point, is the number of risk points.

5. A nest site selection method for non-site rapid evidence collection of traffic for drones according to claim 4, characterized in that, The S3 includes the following sub-steps: S3.1: Calculate and screen out the eligible risk intersections and obtain their coordinates. According to the relevant parameters input by the model, input the relevant data and calculate the coverage radius matrix; Set the parameters of the NSGA-II algorithm, including the population size N, the maximum number of iterations T, and the crossover probability P c , and the mutation probability P m ; S3.2: The chromosome consists of two parts: the nest construction decision segment and the coverage relationship decision segment. The nest construction decision segment uses binary encoding with a length of 2|Q|. The first |Q| bits represent the decision segment X for whether to construct a nest with a capacity of ; the last |Q| bits represent the decision segment X for whether to construct a nest with a capacity of aq ; The coverage relationship decision segment uses binary matrix encoding Z ; the last |Q| bits represent the decision segment X for whether to construct a nest with a capacity of bq ; The coverage relationship decision segment uses binary matrix encoding Z rq , indicating whether the th risk point is covered by the th alternative point. The dimension is R×Q and it needs to satisfy the coverage feasibility constraint. Randomly generate an initial population with a size of N, but allow the chromosomes to temporarily violate the mutual exclusion constraint and the unique coverage constraint; S3.3: For each individual in the population, define a two-objective fitness function: ； ； Among them, is the fitness function of the first target, is the fitness function of the second target; Introduce a multi-constraint violation penalty mechanism to comprehensively handle the following constraints: ； The penalty term is designed as: ； In the formula, , , are penalty coefficients, which are dynamically adjusted according to the constraint priority; Attach the penalty term to the objective function to form a modified two-objective fitness function; ； ； Among them, is the fitness function of the first target after correction, is the fitness function of the first target after correction; it can ensure that individuals violating the constraints are gradually eliminated during the evolution process; S3.4: Perform non-dominated sorting on all individuals in the population to divide the Pareto front levels of the individuals; calculate the crowding degree of individuals within the same front layer; S3.5: Perform genetic operations. The genetic operations are divided into three parts: selection, crossover, and mutation; The selection operation uses the tournament selection method. Randomly select K individuals from the population, compare their non-dominated levels and crowding degrees, and preferentially select individuals with lower front levels. If they are at the same level, select individuals with larger crowding degrees to maintain diversity; The crossover operation performs two-point crossover on the decision segment of the nest construction, randomly selects two crossover points, and exchanges the corresponding segments of the nest decision segment in the parental chromosomes. The coverage relationship segment is dynamically adjusted according to the crossover result to update the coverage relationship matrix of the offspring; ensure that: ； In the mutation operation, with a mutation probability randomly flip a bit in X aq or X aq If the state of a certain nacelle changes after flipping, the corresponding element in the nacelle coverage relationship Z rq must be forced to be set to 0; S3.6: Merge the parental population and the offspring population. The size of the merged population is 2N. Re-perform non-dominated sorting and crowding degree calculation on the merged population; retain the top N optimal individuals in the order of the fronts as the new population; S3.7: If the maximum number of iterations T is reached, terminate the algorithm and output the Pareto front solution set; otherwise, return to S3.4; The Pareto front solution set contains different site selection plan diagrams and their corresponding costs and coverage risk indices considering the penalty coefficient, and generates a coordinate scatter plot that can reflect the two index values; from the Pareto front solution set, according to specific decision-making requirements and preferences, if the decision maker needs to achieve the best coverage efficiency without considering the cost, then select the plan with the highest coverage risk index considering the penalty coefficient in the Pareto front solution; on the contrary, if the decision maker needs the plan with the lowest cost and thus can sacrifice the coverage effect, then select the plan with the lowest cost in the Pareto front solution; if the decision maker needs to balance the two indices for comprehensive consideration, then select other front solutions in the Pareto front solution set, make a decision by referring to the coordinate scatter plot and the site selection plan diagram, so as to select the optimal nest site selection plan.

Images