Unmanned aerial vehicle-oriented nest site selection method for traffic off-site rapid evidence obtaining

Abstract

The invention discloses an unmanned aerial vehicle-oriented nest site selection method for traffic off-site rapid evidence obtaining, and belongs to the technical field of unmanned aerial vehicle nest site selection, and the method comprises the steps: firstly searching for an accident high-incidence intersection, obtaining and processing intersection traffic data in a selected region, carrying out the weight distribution of each index in the data through an entropy weight method, and carrying out the calculation of the weight distribution of each index; calculating a comprehensive score of each intersection by using a TOPSIS method, determining a risk threshold value, determining an accident high-incidence intersection as a risk point and allocating a weight, establishing a rectangular plane coordinate system for the accident risk point obtained in the above step, comparing the risk point to the coordinate system, and obtaining a position coordinate of the risk point; and then establishing a traffic accident non-field evidence obtaining unmanned aerial vehicle nest site selection model, and solving the unmanned aerial vehicle nest site selection model based on an NSGA-II algorithm to obtain an optimal nest site selection scheme.

Description

Technical Field

[0001] The invention belongs to the technical field of drone nest site selection, and in particular relates to a drone nest site selection method for off-site rapid traffic evidence collection. Background Art

[0002] With the rapid increase in the number of motor vehicles in my country, the incidence of traffic accidents is also on the rise. At present, traffic accident evidence collection is mainly done by traffic police arriving at the scene, taking photos of the details of the accident scene, and then making judgments based on the video from the road network cameras. When the accident rate is high, it takes a certain amount of time for traffic police to arrive at the scene, and police resources are limited, which often leads to inefficient accident handling, increased traffic congestion, and even increased risk of secondary accidents.

[0003] In recent years, some cities have tried to introduce drones to assist in evidence collection, which can quickly locate accident scenes and collect video evidence from a high-altitude perspective, alleviating the pressure on ground police forces to a certain extent. However, most existing drone evidence collection models use manual remote control operation. Although they can replace some on-site investigation work, they still require special human control in essence, and cannot break through the bottleneck of human resources, and have certain limitations.

[0004] The drone nest is an automated aviation infrastructure designed specifically for drones. It is equipped with automatic charging stations, data transmission equipment, etc. Drones can automatically take off and land, automatically charge, and automatically maintain themselves in the nest. The nest can collect, store, and analyze data transmitted by drones to provide support for decision-making. In an emergency, the drone nest can quickly deploy drones for aerial inspections or rescue. It is widely used in power inspections, traffic monitoring, security patrols, and other fields. It can replace manual operations, improve work efficiency, and reduce labor costs and safety risks.

[0005] At present, the site selection of drone nests is mostly used for power station line inspections, highway maintenance, etc., and there are fewer nest sites for non-site rapid evidence collection in traffic. In complex traffic scenarios, the site selection of drone nests needs to comprehensively consider factors such as traffic flow and accident-prone areas to ensure that drones can respond quickly and collect evidence accurately. However, existing site selection methods mostly use subjective methods, lack scientific evaluation systems and quantitative indicators, and are 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 method for selecting a nest for rapid off-site traffic evidence collection of unmanned aerial vehicles. The method has a reasonable design, solves the shortcomings of the prior art, and has good effects.

[0007] A method for selecting a nest site for off-site rapid evidence collection of UAV traffic, comprising the following steps: S1: Find intersections with high accident rates; S2: Establish a model for selecting drone nests for off-site evidence collection in traffic accidents; S3: Solve the UAV nest site selection model based on the NSGA-II algorithm to obtain the optimal nest site selection plan.

[0008] Furthermore, the S1 comprises the following sub-steps: S1.1: Obtain and process intersection traffic data in the selected area; S1.2: Use the entropy weight method to assign weights to each indicator in the data; S1.3: Use the TOPSIS method to calculate the comprehensive score of each intersection; S1.4: Establish risk thresholds, identify high-accident intersections as risk points and assign weights; S1.5: For the accident risk points obtained in the above steps, establish a plane rectangular coordinate system, compare the risk points to the coordinate system, and obtain their position coordinates .

[0009] Furthermore, S1.1 is specifically: for all intersections in the selected area, the following data are collected and sorted: The intersection set is defined as: ; in, and Represents negative indicators, namely the number of lanes and sight distance score, represents positive indicators, which are annual number of accidents, daily average traffic volume, average vehicle speed, emergency braking frequency and pedestrian flow; The original data matrix is ​​defined as: ; in, Indicates The intersection is The original value of the indicator; Standardize the collected data: For positive indicators: , ; For negative indicators: , ; in, After standardization The intersection is The value of the indicator, For the The maximum value among the original values ​​of the indicators, For the The minimum value among the original values ​​of the indicators; The S1.2 is specifically: Calculate the index proportion, the expression is: ; in, For the The intersection corresponds to The proportion of indicators, After standardization The intersection is The value of the indicator, ; Calculate the information entropy, the expression is: ; in, For the The information entropy of the indicator, For the The intersection corresponds to The proportion of indicators; when When, define ; Calculate the coefficient of variation, the expression is: ; in, For the The coefficient of variation of the indicators; Determine the weight of each indicator, the expression is: ; in, For the The weight of the indicator, For the The coefficient of variation of the indicators.

[0010] Furthermore, the S1.3 is specifically: constructing a weighting matrix, the elements in the weighting matrix The expression is: ; Define the ideal solution: No. The positive ideal solution expression corresponding to the index is: ; No. The negative ideal solution expression corresponding to the index is: ; Calculate distance: No. The Euclidean distance from the intersection to the positive ideal solution The expression is: ; No. The Euclidean distance from the intersection to the negative ideal solution The expression is: ; Calculate the proximity, the expression is: ;

[0011] in, For the The proximity of each intersection; Output comprehensive score vector : ; Specifically, 1.4 is as follows: for all intersections that have their own comprehensive scores, first determine the threshold setting, which is expressed as: ; in, is the demand sensitivity, The highest rating, is the lowest rating; Using the threshold as the boundary, the intersections with scores higher than the threshold are selected as accident-prone intersections, which are defined as risk points. A weight is assigned to each risk point. , the expression is: ; in, For the The weight of each risk point, For the The risk score of is the number of risk points.

[0012] Furthermore, the S2 comprises the following sub-steps: S2.1: Define the model objective function. The site selection objective is to maximize the coverage of the machine nest and minimize the total cost of the machine nest construction. The expression is: ; ; Each risk point has coordinates , Q is the number of candidate points, indicating the candidate locations of the drone nest, and each candidate point has coordinates ; Use risk points as alternative locations for machine nests to maximize coverage area. is the fixed cost of the machine nest, The capacity is The incremental cost of the nest, The capacity is The incremental cost of the nest, To indicate the Whether the construction capacity of the candidate point is The 0-1 decision variables of the machine nest, To indicate the Whether the construction capacity of the candidate point is The 0-1 variable of the machine nest; To indicate the Is the risk point 0-1 decision variables covering the nest of candidate points; S2.2: Determine the coverage relationship matrix between risk points and alternative points: 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 : ; in, is the maximum communication distance of the UAV, The maximum flight distance within the drone's flight range. Its value follows the following formula: ; in, The longest flight time for drones. The time it takes for a drone to perform a single evidence collection mission. is the flight speed of the drone, is the average flight altitude of the drone; S2.3: Define key model constraints: Coverage feasibility, the expression is: ; To ensure that its coverage meets the coverage distance limit, only when the The machine nest is built at the alternative point and it is connected to the first The distance of the risk point is within the coverage radius, and it is considered as the first risk point. The risk point was The alternative point machine nest is covered; Unique coverage, the expression is: ; Ensure that each risk point is covered by at least one machine nest; Capacity limit, expressed as: ; ensure The machine nest at can meet the range Point demand; Mutually exclusive constraint, the expression is: ; Ensure that only one machine nest is built at each site.

[0013] Furthermore, S3 includes the following sub-steps: S3.1: Calculate and screen out qualified risk intersections and obtain their coordinates, input relevant parameters according to the model, input relevant data and calculate the coverage radius matrix; set NSGA-II algorithm parameters, including population size N, maximum number of iterations T, crossover probability P c , 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 coding with a length of 2|Q|. The first |Q| bits indicate whether to build a capacity of The decision segment of the machine nest X aq The last |Q| bit indicates whether the capacity is built. The decision segment of the machine nest X bq ; The coverage relationship decision segment adopts binary matrix coding Z rq , indicating the Is the risk point The coverage of candidate points is R×Q, and the coverage feasibility constraint must be met; The initial population is randomly generated with a size of N, but chromosomes are allowed to temporarily violate the mutual exclusion constraint and the unique coverage constraint; S3.3: For each individual in the population, define a dual-objective fitness function: ; ; in, is the fitness function of the first objective, is the fitness function of the second objective; A multi-constraint violation penalty mechanism is introduced to comprehensively handle the following constraints: ; The penalty item is designed as: ; In the formula, , , is the penalty coefficient, which is dynamically adjusted according to the constraint priority; The penalty term is added to the objective function to form the modified dual-objective fitness function: ; ; in, is the fitness function of the first objective after correction, is the fitness function of the first objective after correction; 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 and divide the Pareto frontier levels of the individuals; calculate the crowding degree of individuals in the same frontier layer; S3.5: Perform genetic operations, which are divided into three parts: selection, crossover, and mutation; The selection operation adopts the tournament selection method, randomly selecting K individuals from the population, comparing their non-dominated ranks and crowding degrees, and giving priority to individuals with lower frontier ranks. If they are at the same rank, individuals with higher crowding degrees are selected to maintain diversity; The crossover operation performs a two-point crossover on the nest construction decision segment, randomly selects two crossover points, exchanges the corresponding segments of the nest decision segment in the parent chromosome, and dynamically adjusts the coverage relationship segment according to the crossover result, and updates the coverage relationship matrix of the offspring; ensuring that: ; In mutation operation, the mutation probability Randomly flip X aq or X aq If the state of a machine nest changes after flipping, the machine nest needs to cover the relationship Z rq The corresponding elements in are forced to 0; S3.6: Merge the parent population and the child population. The size of the merged population is 2N. Recalculate the non-dominated sorting and crowding degree of the merged population. Keep the top N best individuals as the new population in the frontier order. S3.7: If the maximum number of iterations T is reached, terminate the algorithm and output the Pareto frontier solution set; otherwise return to S3.4; The Pareto frontier solution set contains different site selection scheme diagrams and their corresponding costs and coverage risk indexes considering penalty coefficients, and generates a coordinate scatter plot that can reflect the values ​​of the two indicators; from the Pareto frontier solution set, based on specific decision-making needs 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 frontier solution; conversely, if the decision maker needs the lowest cost plan so that he can give up some coverage effect, then select the lowest cost plan in the Pareto frontier solution; if the decision maker needs to balance the two indicators for comprehensive consideration, then select other frontier solutions in the Pareto frontier solution set, make decisions by comparing the coordinate scatter plot and the site selection scheme diagram, and thus select the optimal machine nest site selection plan.

[0014] Beneficial technical effects brought by the present invention: The present invention proposes a method for drone nest site selection for off-site rapid evidence collection of traffic. After obtaining relevant data of traffic intersections in the area and relevant parameters of drone nests, an accurate and efficient multi-objective optimization drone nest site selection technology can be realized according to regional characteristics. The technology can include two nests with different capacities, and comprehensively considers the cost and coverage effect. Finally, it can output multiple site selections with different indices for reference by decision makers. The technology has good scientificity, reliability and robustness, and can be applied to off-site evidence collection management of traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flow chart of a method for selecting a nest site for rapid off-site traffic evidence collection of unmanned aerial vehicles.

[0016] Figure 2 Schematic diagram of the risk point in the plane rectangular coordinate system in Example 1.

[0017] Figure 3 Schematic diagram of five feasible solutions in the Pareto front solution set in Example 1.

[0018] Figure 4 This is the site selection map for solution 1 (11.74, 3090059.64) in Example 1.

[0019] Figure 5 This is the site selection map for solution 2 (14.73, 4305056.66) in Example 1.

[0020] Figure 6 This is the site selection map for solution 3 (13.74, 3210057.64) in Example 1.

[0021] Figure 7 This is the site selection map for solution 4 (14.43,4275056.96) in Example 1.

[0022] Figure 8 This is the site selection map for solution 5 (15.26,5040056.13) in Example 1.

[0023] Fig. 9 This is a curve chart showing the change of total cost with the number of iterations in Example 1.

[0024] Fig.10 This is a curve showing how the coverage risk index after the comprehensive penalty coefficient changes with the number of iterations in Example 1. DETAILED DESCRIPTION

[0025] The specific implementation of the present invention is further described below in conjunction with specific embodiments: A method for selecting a nest site for off-site rapid evidence collection of UAV traffic, comprising the following steps: S1: Find intersections with high accident rates, including the following sub-steps: S1.1: Obtain and process intersection traffic data in the selected area; For all intersections in the selected area, collect and organize the following data: The intersection set is defined as: ; in, and Represents negative indicators, namely the number of lanes and sight distance score, represents positive indicators, which are annual number of accidents, daily average traffic volume, average vehicle speed, emergency braking frequency and pedestrian flow; The original data matrix is ​​defined as: ; in, Indicates The intersection is The original value of the indicator; The collected data is standardized. The goal of standardization is to map the data to the [0,1] interval to avoid the difference in units or magnitudes of different indicators affecting the fairness of subsequent weight distribution. The rules of standardization are as follows: For positive indicators: , ; For negative indicators: , ; in, After standardization The intersection is The value of the indicator, For the The maximum value among the original values ​​of the indicators, For the The minimum value among the original values ​​of the indicators.

[0026] S1.2: Use the entropy weight method to assign weights to each indicator in the data; The entropy weight method is an objective weighting method based on the principle of information theory. It determines the weight of each indicator by calculating its information entropy. It is often used in multi-indicator comprehensive evaluation and decision analysis. In the present invention, the entropy weight method can be used to assign weights to data in the following four steps: ① Calculate the index proportion, the expression is: ; in, For the The intersection corresponds to The proportion of indicators, After standardization The intersection is The value of the indicator, ; ② Calculate information entropy, the expression is: ; in, For the The information entropy of the indicator, For the The intersection corresponds to The proportion of indicators; when When, define ; ③ Calculate the coefficient of difference, the expression is: ; in, For the The coefficient of variation of the indicators; ④Determine the weight of each indicator, the expression is: ; in, For the The weight of the indicator, For the The coefficient of variation of the indicators.

[0027] S1.3: Use the TOPSIS method to calculate the comprehensive score of each intersection; The TOPSIS method is a multi-criteria decision analysis method that constructs a positive ideal solution (optimal solution) and a negative ideal solution (worst solution) to evaluate the problem, calculates the similarity between each candidate solution and the ideal solution and the negative ideal solution, evaluates and ranks the advantages and disadvantages of each solution, and finally selects the solution closest to the ideal solution as the optimal solution. In the present invention, the TOPSIS method scores each intersection in the following four steps: Construct a weighted matrix, the elements in the weighted matrix The expression is: ; Define the ideal solution: No. The positive ideal solution expression corresponding to the index is: ; No. The negative ideal solution expression corresponding to the index is: ; Calculate distance: No. The Euclidean distance from the intersection to the positive ideal solution The expression is: ; No. The Euclidean distance from the intersection to the negative ideal solution The expression is: ; Calculate the proximity, the expression is: ;

[0028] in, For the The proximity of each intersection; Output comprehensive score vector : .

[0029] S1.4: Establish risk thresholds, identify high-accident intersections as risk points and assign weights, specifically: For all intersections that have their own comprehensive scores, the threshold setting is first determined, and the expression is: ; in, For the required sensitivity, it can be set to α=0.6 and adjusted according to the needs. The highest rating, is the lowest rating; Using the threshold as the boundary, the intersections with scores higher than the threshold are selected as high-accident intersections, which are defined as high-risk points for traffic accidents, and weights are assigned to each risk point. , the expression is: ; in, For the The weight of each risk point, For the The risk score of is the number of risk points.

[0030] S1.5: For the accident risk points obtained in the above steps, establish a plane rectangular coordinate system, compare the risk points to the coordinate system, and obtain their position coordinates .

[0031] S2: Establishing a model for selecting drone nests for off-site evidence collection in traffic accidents, including the following sub-steps: S2.1: Define the model objective function; This model is a multi-objective planning model. The goal of site selection is to maximize the coverage of the machine nest and minimize the total cost of machine nest construction. The expression is: ; ; Each risk point has coordinates , Q is the number of candidate points, indicating the candidate locations of the drone nest, and each candidate point has coordinates ; Use risk points as alternative locations for machine nests to maximize coverage area. is the fixed cost of the machine nest, The capacity is The incremental cost of the nest, The capacity is The incremental cost of the nest, To indicate the Whether the construction capacity of the candidate point is The 0-1 decision variables of the machine nest, To indicate the Whether the construction capacity of the candidate point is The 0-1 variable of the machine nest; To indicate the Is the risk point 0-1 decision variables covering the nest of candidate points; S2.2: Determine the coverage relationship matrix between risk points and alternative points; 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 : ; in, is the maximum communication distance of the UAV, The maximum flight distance within the drone's flight range. Its value follows the following formula: ; in, The longest flight time for drones. The time it takes for a drone to perform a single evidence collection mission. is the flight speed of the drone, is the average flight altitude of the drone; S2.3: Define key model constraints: This model comprehensively considers the UAV's endurance and other related characteristics, and sets the following constraints to ensure that the model meets the needs of actual application scenarios.

[0032] ① Coverage feasibility, the expression is: ; To ensure that its coverage meets the coverage distance limit, only when the The machine nest is built at the alternative point and it is connected to the first The distance of the risk point is within the coverage radius, and it is considered as the first risk point. The risk point was The alternative point machine nest is covered; ②Unique coverage, the expression is: ; Ensure that each risk point is covered by at least one machine nest; ③Capacity limitation, the expression is: ; ensure The machine nest at can meet the range Point demand; ④ Mutually exclusive constraint, the expression is: ; Ensure that only one machine nest is built at each site.

[0033] S3: Solve the drone nest site selection model based on the NSGA-II algorithm to obtain the optimal drone nest site selection plan, including the following sub-steps: S3.1: parameter initialization and data preprocessing; Calculate and screen out qualified risk intersections and obtain their coordinates, input relevant parameters according to the model, input relevant data and calculate the coverage radius matrix; set NSGA-II algorithm parameters, including population size N, maximum number of iterations T, crossover probability , mutation probability ; S3.2: Chromosome encoding and population initialization; The chromosome consists of two parts: the nest construction decision segment and the coverage relationship segment. The nest construction decision segment uses binary coding with a length of 2|Q|. The first |Q| bit indicates whether to build a capacity of The decision segment of the machine nest X aq The last |Q| bit indicates whether the capacity is built. The decision segment of the machine nest X bq ; Covering relationship segment uses binary matrix coding Z rq , indicating the Is the risk point The coverage of candidate points is R×Q, and the coverage feasibility constraint must be met; The initial population is randomly generated with a size of N, but chromosomes are allowed to temporarily violate the mutual exclusion constraint and the unique coverage constraint; S3.3: Fitness function and constraint handling; For each individual in the population, define the dual-objective fitness function: ; ; in, is the fitness function of the first objective, is the fitness function of the second objective; A multi-constraint violation penalty mechanism is introduced to comprehensively handle the following constraints: ; The penalty item is designed as: ; In the formula, , , is the penalty coefficient, which is dynamically adjusted according to the constraint priority; The penalty term is added to the objective function to form the modified dual-objective fitness function: ; ; in, is the fitness function of the first objective after correction, is the fitness function of the first objective after correction; after correction, it can ensure that individuals violating the constraints are gradually eliminated during the evolution process; S3.4: Non-dominated sorting and crowding calculation; Perform non-dominated sorting on all individuals in the population and divide the Pareto frontier levels of the individuals; calculate the crowding degree of individuals in the same frontier layer to ensure population diversity; S3.5: Perform genetic operations, which are divided into three parts: selection, crossover, and mutation; The selection operation adopts the tournament selection method, randomly selecting K individuals from the population, comparing their non-dominated ranks and crowding degrees, and giving priority to individuals with lower frontier ranks. If they are at the same rank, individuals with higher crowding degrees are selected to maintain diversity; The crossover operation performs a two-point crossover on the nest construction decision segment, randomly selects two crossover points, exchanges the corresponding segments of the nest decision segment in the parent chromosome, and dynamically adjusts the coverage relationship segment according to the crossover result, and updates the coverage relationship matrix of the offspring; ensuring that: ; In mutation operation, the mutation probability Randomly flip X aq or X aqIf the state of a machine nest changes after flipping, the machine nest needs to cover the relationship Z rq The corresponding elements in are forced to 0; S3.6: Population update; Merge the parent population and the child population. The size of the merged population is 2N. Recalculate the non-dominated sort and crowding degree of the merged population. Keep the first N best individuals as the new population in the frontier order (if the total number of individuals on a certain frontier exceeds the remaining capacity, select from high to low according to the crowding degree); S3.7: Termination decision and Pareto frontier solution set output; If the maximum number of iterations T is reached, the algorithm is terminated and the Pareto frontier solution set is output, otherwise it returns to S3.4; The Pareto frontier solution set contains different site selection scheme diagrams and their corresponding costs and coverage risk indexes considering penalty coefficients, and generates a coordinate scatter plot that can reflect the values ​​of the two indicators; from the Pareto frontier solution set, according to specific decision-making needs 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 frontier solution; conversely, if the decision maker needs the lowest cost plan so that he can give up some coverage effect, then select the lowest cost plan in the Pareto frontier solution; if the decision maker needs to balance the two indicators for comprehensive consideration, then select other frontier solutions in the Pareto frontier solution set, and make decisions by comparing the coordinate scatter plot and the site selection scheme diagram, so as to select the optimal machine nest site selection plan.

[0034] Embodiment 1: In this embodiment, a certain area of ​​3000m*3000m in a certain city is obtained, and the required indicators of all intersections in it are obtained. After standardization, entropy weighting and TOPSIS scoring, the demand sensitivity is set. =0.6, after comprehensive calculation, 146 intersections exceeding the threshold were selected and weighted accordingly. Set it as the risk point and plot it in the plane rectangular coordinate system as follows Figure 2 shown.

[0035] Setting fixed costs for drone nests = 80,000 yuan, incremental cost = 10,000 yuan, incremental cost = 25,000 yuan; the capacities of the two machine nests are = 1, = 2; UAV maximum communication distance = 3000m, maximum endurance time = 30min, the time required to perform a single evidence collection task = 8min, flight speed = 200m / min, flight altitude = 200m.

[0036] Program the NSGA-Ⅱ algorithm used in this model based on the Pycharm platform in the Python 3.12 environment, input relevant parameters, and set the penalty coefficient = = = 1, population size N = 500, maximum number of iterations T = 600, crossover probability = 0.75, mutation probability = 0.05.

[0037] Running the algorithm finally gets the Pareto front solution set as follows Figure 3 As shown in , it can be seen that there are five feasible solutions after deduplication, and the corresponding total cost and coverage risk index after comprehensive penalty coefficient are marked at each solution. The location diagram corresponding to each solution is shown in Figure 4-8 As shown in , these site selection maps include the construction reference locations of the two machine nests and the locations of all risk points, and also mark the cost and coverage risk index corresponding to the solution. At the same time, the total cost change curve with the number of iterations is also output as shown in Fig. 9 As shown in the figure; the coverage risk index after the comprehensive penalty coefficient changes with the number of iterations as shown in the figure Fig.10 shown.

[0038] 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 technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for selecting a nest site for rapid off-site traffic evidence collection of unmanned aerial vehicles, characterized in that: The following steps are involved: S1: Find intersections with high accident rates; S2: Establish a model for selecting drone nests for off-site evidence collection in traffic accidents; S3: Solve the UAV nest site selection model based on the NSGA-II algorithm to obtain the optimal nest site selection plan.

2. The method for selecting a nest site for rapid off-site traffic evidence collection for UAVs according to claim 1 is characterized in that: The S1 comprises the following sub-steps: S1.1: Obtain and process intersection traffic data in the selected area; S1.2: Use the entropy weight method to assign weights to each indicator in the data; S1.3: Use the TOPSIS method to calculate the comprehensive score of each intersection; S1.4: Establish risk thresholds, identify high-accident intersections as risk points and assign weights; S1.5: For the accident risk points obtained in the above steps, establish a plane rectangular coordinate system, compare the risk points to the coordinate system, and obtain their position coordinates .

3. The method for selecting a nest site for rapid off-site traffic evidence collection for UAVs according to claim 2 is characterized in that: S1.1 specifically includes: for all intersections in the selected area, collect and organize the following data: The intersection set is defined as: ； in, and Represents negative indicators, namely the number of lanes and sight distance score, represents positive indicators, which are annual number of accidents, daily average traffic volume, average vehicle speed, emergency braking frequency and pedestrian flow; The original data matrix is ​​defined as: ； in, Indicates The intersection is The original value of the indicator; Standardize the collected data: For positive indicators: ， ； For negative indicators: ， ； in, After standardization The intersection is The value of the indicator, For the The maximum value among the original values ​​of the indicators, For the The minimum value among the original values ​​of the indicators; The S1.2 is specifically: Calculate the index proportion, the expression is: ； in, For the The intersection corresponds to The proportion of indicators, After standardization The intersection is The value of the indicator, ; Calculate the information entropy, the expression is: ； in, For the The information entropy of the indicator, For the The intersection corresponds to The proportion of indicators; when When, define ; Calculate the coefficient of variation, the expression is: ； in, For the The coefficient of variation of the indicators; Determine the weight of each indicator, the expression is: ； in, For the The weight of the indicator, For the The coefficient of variation of the indicators.

4. The method for selecting a nest site for rapid off-site traffic evidence collection for UAVs according to claim 3 is characterized in that: The S1.3 is specifically: constructing a weighting matrix, the elements in the weighting matrix The expression is: ； Define the ideal solution: No. The positive ideal solution expression corresponding to the index is: ； No. The negative ideal solution expression corresponding to the index is: ； Calculate distance: No. The Euclidean distance from the intersection to the positive ideal solution The expression is: ； No. The Euclidean distance from the intersection to the negative ideal solution The expression is: ； Calculate the proximity, the expression is: ； in, For the The proximity of each intersection; Output comprehensive score vector : ； Specifically, 1.4 is as follows: for all intersections that have their own comprehensive scores, first determine the threshold setting, which is expressed as: ； in, is the demand sensitivity, The highest rating, is the lowest rating; Using the threshold as the boundary, the intersections with scores higher than the threshold are selected as accident-prone intersections, which are defined as risk points. A weight is assigned to each risk point. , the expression is: ； in, For the The weight of each risk point, For the The risk score of is the number of risk points.

5. The method for selecting a nest site for rapid off-site traffic evidence collection for UAVs according to claim 4 is characterized in that: The S2 comprises the following sub-steps: S2.1: Define the model objective function. The site selection objective is to maximize the coverage of the machine nest and minimize the total cost of the machine nest construction. The expression is: ； ； Each risk point has coordinates , Q is the number of candidate points, indicating the candidate locations of the drone nest, and each candidate point has coordinates ; Use risk points as alternative locations for machine nests to maximize coverage area. is the fixed cost of the machine nest, The capacity is The incremental cost of the nest, The capacity is The incremental cost of the nest, To indicate the Whether the construction capacity of the candidate point is The 0-1 decision variables of the machine nest, To indicate the Whether the construction capacity of the candidate point is The 0-1 variable of the machine nest; To indicate the Is the risk point 0-1 decision variables covering the nest of candidate points; S2.2: Determine the coverage relationship matrix between risk points and alternative points: 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 : ； in, is the maximum communication distance of the UAV, The maximum flight distance within the drone's flight range. Its value follows the following formula: ； in, The longest flight time for drones. The time it takes for a drone to perform a single evidence collection mission. is the flight speed of the drone, is the average flight altitude of the drone; S2.3: Define key model constraints: Coverage feasibility, the expression is: ； To ensure that its coverage meets the coverage distance limit, only when the The machine nest is built at the alternative point and it is connected to the first The distance of the risk point is within the coverage radius, and it is considered as the first risk point. The risk point was The alternative point machine nest is covered; Unique coverage, the expression is: ； Ensure that each risk point is covered by at least one machine nest; Capacity limit, expressed as: ； ensure The machine nest at can meet the range Point demand; Mutually exclusive constraint, the expression is: ； Ensure that only one machine nest is built at each site.

6. The method for selecting a nest site for rapid off-site traffic evidence collection for UAVs according to claim 5 is characterized in that: The S3 comprises the following sub-steps: S3.1: Calculate and screen out the risk intersections that meet the conditions and obtain their coordinates, input relevant parameters according to the model, input relevant data and calculate the coverage radius matrix; Set the NSGA-II algorithm parameters, including population size N, maximum number of iterations T, and crossover probability P c , 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 coding with a length of 2|Q|. The first |Q| bits indicate whether to build a capacity of The decision segment of the machine nest X aq The last |Q| bit indicates whether the capacity is built. The decision segment of the machine nest X bq ; The coverage relationship decision segment adopts binary matrix coding Z rq , indicating the Is the risk point The coverage of candidate points is R×Q, and the coverage feasibility constraint must be met; The initial population is randomly generated with a size of N, but chromosomes are allowed to temporarily violate the mutual exclusion constraint and the unique coverage constraint; S3.3: For each individual in the population, define a dual-objective fitness function: ； ； in, is the fitness function of the first objective, is the fitness function of the second objective; A multi-constraint violation penalty mechanism is introduced to comprehensively handle the following constraints: ； The penalty item is designed as: ； In the formula, , , is the penalty coefficient, which is dynamically adjusted according to the constraint priority; The penalty term is added to the objective function to form the modified dual-objective fitness function: ； ； in, is the fitness function of the first objective after correction, is the fitness function of the first objective after correction; 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 and divide the Pareto frontier levels of the individuals; calculate the crowding degree of individuals in the same frontier layer; S3.5: Perform genetic operations, which are divided into three parts: selection, crossover, and mutation; The selection operation adopts the tournament selection method, randomly selecting K individuals from the population, comparing their non-dominated ranks and crowding degrees, and giving priority to individuals with lower frontier ranks. If they are at the same rank, individuals with higher crowding degrees are selected to maintain diversity; The crossover operation performs a two-point crossover on the nest construction decision segment, randomly selects two crossover points, exchanges the corresponding segments of the nest decision segment in the parent chromosome, and dynamically adjusts the coverage relationship segment according to the crossover result, and updates the coverage relationship matrix of the offspring; ensuring that: ； In mutation operation, the mutation probability Randomly flip X aq or X aq If the state of a nest changes after flipping, the nest needs to cover the relation Z rq The corresponding elements in are forced to 0; S3.6: Merge the parent population and the child population. The size of the merged population is 2N. Recalculate the non-dominated sorting and crowding degree of the merged population. Keep the top N best individuals as the new population in the frontier order. S3.7: If the maximum number of iterations T is reached, terminate the algorithm and output the Pareto frontier solution set; otherwise return to S3.4; The Pareto frontier solution set contains different site selection scheme diagrams and their corresponding costs and coverage risk indexes considering penalty coefficients, and generates a coordinate scatter plot that can reflect the values ​​of the two indicators; from the Pareto frontier solution set, based on specific decision-making needs 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 frontier solution; conversely, if the decision maker needs the lowest cost plan so that he can give up some coverage effect, then select the lowest cost plan in the Pareto frontier solution; if the decision maker needs to balance the two indicators for comprehensive consideration, then select other frontier solutions in the Pareto frontier solution set, make decisions by comparing the coordinate scatter plot and the site selection scheme diagram, and thus select the optimal machine nest site selection plan.

Images