A path optimization method for multi-UAV road network traffic flow monitoring

Abstract

The present invention provides a method for optimizing the path of traffic flow monitoring of a multi-UAV road network, comprising the steps of: S1, collecting road network information data and monitoring requirements, and constructing a multi-UAV traffic flow monitoring road network; S2, constructing a multi-objective optimization model for the path planning of traffic flow monitoring of a multi-UAV road network under the constraint of collection interval; S3, solving the multi-objective optimization model using a variable neighborhood local search genetic algorithm, i.e., a meme algorithm, constructed based on feasible solutions, to obtain a set of optimal paths for UAVs. The present invention constructs multiple optimization target models for minimizing equipment cost and minimizing energy consumption, and solves them through a variable neighborhood local search genetic algorithm, thereby obtaining a set of optimal path planning for multiple UAVs under the constraints of the monitored accuracy of road network traffic flow and the monitored collection interval market, providing economical and portable strong guarantees for the real-time and safety of road traffic informatization.

Description

Technical Field

[0001] The invention relates to a multi-UAV path planning method for monitoring traffic flow, and belongs to the technical field of traffic planning. Background Art

[0002] Drone monitoring is a new type of monitoring equipment. Its portability and flexibility can solve the areas that traditional monitoring equipment cannot cover and the advantages of real-time and accuracy that cannot be achieved. At the same time, the reasonable path planning and scheduling scheme of drones can greatly save equipment costs and energy losses, providing more powerful support for more efficient promotion of drone road network traffic flow monitoring.

[0003] However, in the actual UAV path planning process, due to the need to consider factors such as the coordinated scheduling of multiple UAVs, the limited endurance of UAVs, and the need to monitor the road network with collection intervals, the multi-UAV path planning has problems such as low coordination efficiency and difficulty in obtaining feasible solutions. At the same time, the optimization process that only considers monitoring efficiency with the goal of shortening the minimum service time does not meet the feasibility requirements of the problem, and it is urgent to design a multi-objective optimization solution with the goal of completing the monitoring task cost.

[0004] At present, most drone path planning is used for "last mile" logistics distribution, and the optimization content is only the order of drone's visit to nodes to obtain the shortest path or the shortest task completion time. As a variant of traditional VRP, this method lacks research depth in the field of traffic flow monitoring. When faced with the problem of limited drone path plan, the traditional VRP model optimization idea is difficult to solve the arc path task requirements.

[0005] Therefore, an efficient solution method is proposed for the optimization problem of multi-UAV road network traffic flow monitoring path under the constraint of road section collection interval, so as to reduce the equipment hardware cost and energy consumption cost, improve the quality and efficiency of task completion, and promote the development of new traffic flow monitoring methods and the efficiency of urban road traffic management departments. In summary, considering the accuracy of UAV monitoring and the time interval constraints of the monitored road network, the UAV scheduling plan and the path planning of each UAV trip are reasonably selected to determine the optimal global path planning plan. Ultimately, the purpose of reducing the equipment cost of the number of UAVs and the energy consumption cost of UAV charging under the premise of completing the monitoring task requirements is achieved, and the degree of traffic flow information mastery by urban road traffic management departments is improved, thereby improving the quality of road network traffic. Summary of the invention

[0006] In response to the above problems to be solved, the present invention provides a method for determining the optimal global multi-UAV multi-trip path planning scheme that takes into account UAV endurance, UAV road network monitoring accuracy, and time interval constraints for monitored road sections, so as to meet the need for considering collection time intervals for multiple monitored road sections in the road network; and provides a multi-UAV path optimization method that provides different path optimization schemes for urban road management departments according to different needs.

[0007] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0008] The present invention provides a method for optimizing the path of multi-UAV road network traffic flow monitoring, comprising the following steps:

[0009] S1. Construct a multi-UAV traffic flow monitoring road network: Construct a road network G based on the traffic flow monitoring needs through road network data collection, G = (V, C, E, A), where V is the set of road network intersections, C is the set of charging stations distributed at some road network intersections, E is the set of road network sections, and A is the set of road sections that need to be monitored;

[0010] S2. Construct a multi-objective optimization model: Construct a multi-objective optimization model for path planning of multi-UAV road network traffic flow monitoring under the constraint of collection interval. The multi-objective optimization model includes two objectives, namely, minimizing the equipment cost minf1 and minimizing the energy consumption minf2, where f1 is the equipment cost and f2 is the energy consumption;

[0011] S3. Solve the multi-objective optimization model to obtain the optimal path set of the UAV: ​​Use the evolutionary algorithm based on the variable neighborhood local search strategy, that is, the meme algorithm, to solve the multi-objective optimization model and obtain the optimal path set of the UAV, specifically including:

[0012] S31, algorithm setting parameters: including algorithm crossover probability p c , mutation probability p c and the maximum number of iterations;

[0013] S32, encoding: encoding the chromosome, wherein the length of the chromosome represents the number of road sections with monitoring requirements, and the value of the chromosome represents the number of drones serving the specific road section;

[0014] S33, generating an initial population: according to the encoding method, in the decision search space of the multi-objective optimization model, the initial path is constructed by using the time interval constraint to satisfy the step-by-step review method, that is, a path is randomly selected from the paths that are allowed to fly under the premise of satisfying the endurance of the drone, and the initial chromosome is obtained. In this way, a population containing N individuals is gradually generated;

[0015] S34. Genetic algorithm is used in combination with variable neighborhood search strategy to generate offspring near the initial feasible solution population, and excellent offspring with better objective function values ​​are retained through environmental selection strategy considering the optimization effect of objective function, and finally the optimal UAV flight path set is obtained through continuous iteration.

[0016] In the present invention, genetic algorithm and evolutionary algorithm can be regarded as equivalent concepts, and memetic algorithm is a kind of genetic algorithm.

[0017] In a specific implementation, the two objective functions described in step S2 are as follows:

[0018]

[0019]

[0020] Among them, f1 is the equipment cost, and the equipment includes drones; f2 is the total energy consumption, that is, energy consumption; V is the set of road network intersections, that is, the node set; A is the set of arc paths with monitoring requirements; u is the set of drones; v is the horizontal flight speed of the drone; T is the complete road network traffic flow monitoring cycle; t u is the longest flight time of the drone under fixed power consumption; t c How long it takes to charge the drone to completion; is the total number of visits to the road segment (i, j) by UAV k in the actual path planning results, 0 means that UAV k has not visited the (i, j) section in the global monitoring task; is the flight distance of UAV k visiting road section (i, j); Δh i and Δv i are the altitude adjustment value of the UAV at intersection i and the corresponding altitude adjustment speed of ascent or descent; t a and t p are the number of monitoring times and the time of hovering and taking photos on the road sections with monitoring tasks in the actual path planning results of the UAV; h a is the flight altitude of the drone on the road section a with monitoring requirements; va represents the visual angle of the drone; d a is the length of section a.

[0021] In a specific implementation, the two objective functions include the following constraints:

[0022] 1) UAV monitoring accuracy constraint related to flight altitude: f(h)>Acc

[0023] Where f(h) is the functional relationship between the UAV flight altitude and monitoring accuracy, which needs to be obtained by combining the data analysis and fitting of real scene collection; where Acc is the accuracy requirement, which is a constant;

[0024] 2) The difference in take-off and landing heights between two adjacent sections of the road to be monitored: Δh a =h a -h a-1 ,a,a-1∈A

[0025] 3) Drone endurance constraints:

[0026] in is the remaining flight time of the kth UAV after the monitoring section a ends; Same as d a Consistent, also refers to the length of section a; It is the ratio between the difference between the monitoring height of the previous section and the monitoring height of the next section and the speed of ascent or descent, and the speed of ascent or descent is determined by the positive or negative height difference; Δv is the descent speed to the ground before the battery is exhausted;

[0027] 4) The road section is monitored at a certain time interval:

[0028] in is the collection time interval constraint of the monitored road section a, is the time when road section a is monitored for the nth time, is the time when road section a is monitored for the n+1th time, T is the duration of the monitoring task in the entire time period, t a It is the number of monitoring times of the road section a with monitoring tasks in the actual path planning result of the UAV.

[0029] In a specific embodiment, the multi-objective optimization model includes the following assumptions:

[0030] The road network for multi-UAV traffic flow monitoring is fixed according to the problem requirements; the monitored time intervals of some sections to be monitored are allowed to be different but always remain unchanged; the power consumption of the UAV during ascent, descent and flight is the same, and only the speed is different; the functional relationship between the UAV flight altitude and the monitoring accuracy is obtained through data analysis and fitting under a large number of different time periods and different environmental conditions, so the monitoring accuracy is no longer affected by the environment; the number and layout of charging stations absolutely meet the requirements for UAVs to charge after performing part of the mission.

[0031] In a specific embodiment, the road network data includes the starting and ending point coordinates of the monitored sections, the connectivity of the non-monitored sections, the intersection coordinates of all sections of the road network, the collection time interval of all sections to be monitored, the flight altitude of the drone in each monitored section, the layout of the charging stations, and the charging time.

[0032] In a specific implementation, the step S34 specifically includes:

[0033] S341, in the process of constructing solutions, the constraint satisfaction is considered, and a population of N feasible solutions is constructed as the initial population;

[0034] S342, calculating the objective function value: calculating f1 and f2 of each individual in the initial population 1 and the initial population 2;

[0035] S343, using the f1 and f2 objective function values ​​of all individuals in the population to perform Pareto sorting to obtain individuals with non-dominated objective functions, using the binary tournament selection method to select two excellent parent individuals from the non-dominated individuals to perform crossover operations to generate temporary offspring and perform mutation operations on randomly selected individuals from the temporary offspring, and finally using a local search strategy to guide the offspring individuals to improve their satisfaction of constraints;

[0036] S344, repeat the above genetic operation until N offspring individuals are generated, and merge the offspring and parent individuals into 2N sets, and then use Pareto non-dominated sorting to retain the N individuals with the best convergence and distribution to form the next generation population;

[0037] S345, repeat the algorithm process of S341-S344 until the algorithm reaches the maximum number of iterations and terminates;

[0038] S346, decoding the obtained optimal solution, and exporting and displaying the obtained optimal path set.

[0039] In a specific implementation, the crossover operation on the parent individuals in step S343 is as follows: after pairing, the parent individuals randomly selected from the parent population through Pareto sorting and binary tournament have their sequences A and B of equal length in the two individuals swapped in order from the same corresponding positions.

[0040] In a specific implementation, step S343 performs the mutation operation on the temporary offspring individuals as follows: after performing the crossover operation on the parent individuals, randomly select an individual from the temporary offspring of the parent population, randomly generate two numbers as mutation positions, swap the information on the two mutation positions of the randomly selected individual, and perform the mutation operation on all other temporary offspring individuals in this way.

[0041] In a specific implementation, step S343 operates the local search strategy after the offspring is generated as follows: after performing genetic operations including crossover and mutation on the parent individuals of the population, in order to avoid the genetic operations causing the offspring solution to lose the feasibility of the original parent solution, a local search strategy is used to change the genes of the offspring individuals so that they have a higher similarity with the parent individuals to guide the offspring individuals to improve their feasible probability.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention constructs a multi-objective optimization model for minimizing equipment cost (number of drones) and energy cost (number of charging times) during the traffic flow monitoring process of a multi-UAV road network, and solves it through a variety of evolutionary algorithms to obtain multiple optimal path sets, which meets the monitoring needs of the road network considering the collection interval constraints while providing the optimal cost path planning solution.

[0044] 2. The present invention will adopt a feasible solution construction method to reduce the feasibility difficulty of initializing solutions to complex problems, improve problem solving efficiency and final solution quality.

[0045] 3. The essence of the problem involved in the present invention is the multi-trip constrained arc path planning problem. The model and solution method of the problem can not only be applied to the UAV monitoring path planning task, but also can be extended to coastal patrol, fire inspection, takeaway delivery and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of the present invention.

[0047] Figure 2 The following is a schematic diagram of an actual route including the road section monitoring requirements and the drone path planning in an embodiment of the present invention. Among them, Figure A is a schematic diagram of the road section monitoring requirements in the present invention, and Figure B is a schematic diagram of the actual route of the drone path planning in the present invention.

[0048] Figure 3 Schematic diagram of encoding of a feasible solution construction of a genetic algorithm according to an embodiment of the present invention.

[0049] Figure 4 for Figure 3 Path planning scheme corresponding to a specific example under encoding situation.

[0050] Figure 5 This is the genetic algorithm evolution framework diagram of the present invention.

[0051] Figure 6 This is a flow chart of the genetic algorithm solution and solution decoding of the present invention. DETAILED DESCRIPTION

[0052] The present invention will be described in detail below in combination with the embodiments and the accompanying drawings. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0053] Figure 1 This is a flow chart of the present invention, which specifically includes steps S1 to S3.

[0054] Figure 2The following is a schematic diagram of an actual route including the road section monitoring requirements and the drone path planning in an embodiment of the present invention. Figure A is a schematic diagram of the road section monitoring requirements in the present invention, which corresponds to step S1 in the present invention; Figure B is a schematic diagram of the actual route of the drone path planning in the present invention, which corresponds to step S3 in the present invention. Figure 2 In A, both green and orange represent "monitoring demand sections", but these two different colors represent different intervals.

[0055] Figure 3 Schematic diagram of a feasible solution construction coding of a genetic algorithm according to an embodiment of the present invention, which corresponds to step S32 in the present invention.

[0056] Figure 4 for Figure 3 The path planning scheme corresponding to the specific sample under the coding situation corresponds to step S32 in the present invention.

[0057] Figure 5 The genetic algorithm evolution framework diagram of the present invention corresponds to step S34 of the present invention. t " corresponds to step S341; "calculate individual fitness" corresponds to step S342; "select, cross, local search" corresponds to step S343; "evaluate multiple local search temporary offspring and select the best to generate a new generation of population P t+1 " corresponds to step S344; "output final population P" corresponds to step S345.

[0058] Figure 6 This is a flow chart of the genetic algorithm solution and solution decoding of the present invention, which corresponds to step S3 of the present invention.

[0059] Specifically, the present invention provides a method for optimizing the path of multi-UAV road network traffic flow monitoring, comprising the following steps:

[0060] S1. Collect traffic flow monitoring requirements to build a road network G, G = (V, C, E, A), where V is a set of road network intersections, C is a set of charging stations distributed at some road network intersections, E is a set of road network sections, and A is a set of road sections that need to be monitored. Preferably, the road network data includes the coordinates of the starting and ending points of the road sections to be monitored, the communication status of the non-monitored sections, the intersection coordinates of all sections of the road network, the collection time interval of all sections to be monitored, the flight altitude of the drone in each monitored section, and the layout of the charging stations.

[0061] S2. Construct a multi-objective optimization model for path planning of multi-UAV road network traffic flow monitoring under collection interval constraints. The multi-objective optimization model includes two objectives, namely, minimizing equipment cost minf1 and minimizing energy consumption minf2, where f1 is equipment cost and f2 is energy consumption;

[0062] Specifically, the two objective functions described in step S2 are as follows:

[0063]

[0064]

[0065] Among them, f1 is the equipment cost (number of drones), f2 is the total energy consumption; V is the set of intersections (nodes) in the road network; A is the set of arc paths with monitoring requirements; u is the set of drones; v is the horizontal flight speed of the drone; T is the complete road network traffic flow monitoring cycle; t u is the longest flight time of the drone under fixed power consumption; t c How long it takes to charge the drone to completion; is the total number of visits to the road segment (i, j) by UAV k in the actual path planning results, and 0 means that UAV k has not visited the road segment (i, j) in the global monitoring task; is the flight distance of UAV k visiting road section (i, j); Δh i and Δv i are the altitude adjustment value of the UAV at intersection i and the corresponding altitude adjustment speed of ascent or descent; t a and t p are the number of monitoring times and the time of hovering and taking photos on the road sections with monitoring tasks in the actual path planning results of the UAV; h a is the flying height of the UAV on the road section a where monitoring is required; va represents the visual angle of the UAV.

[0066] Preferably, the two objective functions include the following constraints:

[0067] 1) UAV monitoring accuracy constraint (related to flight altitude): f(h)>Acc

[0068] Where f(h) is the functional relationship between the UAV flight altitude and monitoring accuracy, which needs to be analyzed and fitted in combination with the data collected in the real scene;

[0069] 2) The difference in take-off and landing heights between two adjacent sections of the road to be monitored: Δh a =h a -h a-1 ,a,a-1∈A

[0070] 3) Drone endurance constraints:

[0071] in is the remaining flight time of the kth UAV after the monitoring section a ends;

[0072] 4) The road section is monitored at a certain time interval:

[0073] in is the collection time interval constraint of the monitored road section a, is the time when road section a is monitored for the nth time, T is the duration of the monitoring task in the entire time period, and t a It is the number of monitoring times of the road section a with monitoring tasks in the actual path planning result of the UAV.

[0074] The road network for multi-UAV traffic flow monitoring is fixed according to the problem requirements; the monitored time intervals of some sections to be monitored are allowed to be different but always remain unchanged; the power consumption of the UAV during ascent, descent and flight is the same, and only the speed is different; the functional relationship between the UAV flight altitude and the monitoring accuracy is obtained through data analysis and fitting under a large number of different time periods and different environmental conditions, so the monitoring accuracy is no longer affected by the environment; the number and layout of charging stations absolutely meet the requirements for UAVs to charge after performing part of the mission.

[0075] S3. Using the evolutionary algorithm based on the variable neighborhood local search strategy, namely the meme algorithm, the multi-objective optimization model is solved to obtain the optimal path set of the UAV, such as Figure 6 As shown, specifically including:

[0076] S31, algorithm setting parameters: including algorithm crossover probability p c , mutation probability p c and the maximum number of iterations;

[0077] S32, Encoding: Encode the chromosome, such as Figure 3 As shown in the figure, the chromosome is encoded, where the length of the chromosome represents the number of sections with monitoring requirements, and the value of the chromosome represents the number of drones serving a specific section; the chromosome is divided into |A| groups, |A| is the number of sections with monitoring requirements. The number of gene bits in each group represents the number of times task arc a is visited The upper limit of the chromosome value is the upper limit of the number of drones. Figure 3 The actual solution given by the encoding method shown is as follows Figure 4 As shown, the actual path access sequence after decoding of the scheme and the path planning result of the UAV are given;

[0078] S33, generating an initial population: according to the encoding method, in the decision space, the initial solution path access sequence is constructed by using the constraint satisfaction step-by-step review method, and the solution access sequence is randomly selected from the next feasible section for construction to obtain an initial solution, and the construction process is repeated N times to obtain an initial population, where the number of individuals in the population is N, and N is 100;

[0079] S34. Genetic algorithm is used in combination with variable neighborhood search strategy to generate offspring near the initial feasible solution population, and excellent offspring with better objective function values ​​are retained through environmental selection strategy considering the optimization effect of objective function, and finally the optimal UAV flight path set is obtained through continuous iteration.

[0080] S341, in the process of constructing solutions, the constraint satisfaction is considered, and a population of N feasible solutions is constructed as the initial population;

[0081] S342, calculate the objective function value: calculate f1 and f2 of each individual in the initial population 1 and the initial population 2. Considering various constraints, the calculation of the objective function value is related to the constraint violation value. Since this problem is a minimization optimization problem, the infeasible solution can be eliminated by increasing the objective value through the penalty function. Specifically, if the individual is an infeasible solution, the objective function value is equal to the constraint violation value of the solution plus a large penalty value. Otherwise, the individual is a feasible solution and the objective function value remains unchanged.

[0082] S343, using the f1 and f2 objective function values ​​of all individuals in the population to perform Pareto sorting to obtain individuals with non-dominated objective functions, using the binary tournament selection method to select two excellent parent individuals from the non-dominated individuals to perform crossover operations to generate temporary offspring and perform mutation operations on randomly selected individuals from the temporary offspring, and finally using a local search strategy to guide the offspring individuals to improve their satisfaction of constraints;

[0083] Specifically, step S343 performs the crossover operation on the parent individuals as follows: after pairing, the parent individuals randomly selected from the parent population through Pareto sorting and binary tournament have their sequences A and B of equal length in the two individuals exchanged in order from the same corresponding positions.

[0084] Specifically, step S343 performs the mutation operation on the temporary offspring individuals as follows: after performing the crossover operation on the parent individuals, randomly select an individual from the temporary offspring of the parent population, randomly generate two numbers as the mutation positions, swap the information on the two mutation positions of the randomly selected individual, and perform the mutation operation on all other temporary offspring individuals in this way.

[0085] Specifically, step S343 operates the local search strategy after the offspring is generated as follows: after the genetic operation (crossover, mutation) is completed on the parent individuals of the population, in order to avoid the genetic operation causing the offspring solution to lose the feasibility of the original parent solution, a local search strategy will be used to change the genes of the offspring individuals to make them have a higher similarity with the parent individuals to guide the offspring individuals to improve their feasible probability.

[0086] S344, repeat the above genetic operation until N offspring individuals are generated, and merge the offspring and parent individuals into 2N sets, and then use Pareto non-dominated sorting to retain the N individuals with the best convergence and distribution to form the next generation population;

[0087] S345, repeat the algorithm process of S341-S344 until the algorithm reaches the maximum number of iterations and terminates;

[0088] This includes using a constraint processing method based on the extended Pareto dominance relationship to optimize the population after determining the search structure and search strategy, and obtaining the optimal path set to balance the convergence and diversity of the optimal solution set.

[0089] S346, decoding the obtained optimal solution, and exporting and displaying the obtained optimal path set.

[0090] In general, the present invention provides a method for optimizing the path of multi-UAV road network traffic flow monitoring, comprising the steps of: S1, collecting road network information data and monitoring requirements, and constructing a multi-UAV traffic flow monitoring road network; S2, constructing a multi-objective optimization model for multi-UAV road network traffic flow monitoring path planning under collection interval constraints; S3, using a variable neighborhood local search genetic algorithm, i.e., a meme algorithm, constructed based on feasible solutions to solve the multi-objective optimization model and obtain an optimal path set for UAVs.

[0091] The present invention constructs multiple optimization target models for minimizing equipment costs and minimizing energy consumption, and solves them through a variable neighborhood local search genetic algorithm, thereby obtaining a set of optimal path planning for multiple drones while satisfying the market constraints of road network traffic flow monitoring accuracy and monitored collection interval, providing economical and portable strong guarantees for the real-time and safety of road traffic informatization.

[0092] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for optimizing the path of multi-UAV road network traffic flow monitoring, characterized in that: The following steps are involved: S1. Construct a multi-UAV traffic flow monitoring road network: Construct a road network G based on the traffic flow monitoring needs through road network data collection, G = (V, C, E, A), where V is the set of road network intersections, C is the set of charging stations distributed at some road network intersections, E is the set of road network sections, and A is the set of road sections that need to be monitored; S2. Construct a multi-objective optimization model: Construct a multi-objective optimization model for path planning of multi-UAV road network traffic flow monitoring under the constraint of collection interval. The multi-objective optimization model includes two objectives, namely, minimizing the equipment cost minf1 and minimizing the energy consumption minf2, where f1 is the equipment cost and f2 is the energy consumption; S3. Solve the multi-objective optimization model to obtain the optimal path set of the UAV: ​​Use the evolutionary algorithm based on the variable neighborhood local search strategy, that is, the meme algorithm, to solve the multi-objective optimization model and obtain the optimal path set of the UAV, specifically including: S31, algorithm setting parameters: including algorithm crossover probability p c , mutation probability p c and the maximum number of iterations; S32, encoding: encoding the chromosome, wherein the length of the chromosome represents the number of road sections with monitoring requirements, and the value of the chromosome represents the number of drones serving the specific road section; S33, generating an initial population: according to the encoding method, in the decision search space of the multi-objective optimization model, the initial path is constructed by using the time interval constraint to satisfy the step-by-step review method, that is, a path is randomly selected from the paths that are allowed to fly under the premise of satisfying the endurance of the drone, and the initial chromosome is obtained. In this way, a population containing N individuals is gradually generated; S34, using genetic algorithm combined with variable neighborhood search strategy to generate offspring near the initial feasible solution population, and retaining excellent offspring with better objective function values ​​through environmental selection strategy considering the optimization effect of objective function, and finally continuously iterating to obtain the optimal UAV flight path set; The two objective functions described in step S2 are as follows: Among them, f1 is the equipment cost, and the equipment includes drones; f2 is the total energy consumption, that is, energy consumption; V is the set of road network intersections, that is, the node set; A is the set of arc paths with monitoring requirements; u is the set of drones; v is the horizontal flight speed of the drone; T is the complete road network traffic flow monitoring cycle; t u is the longest flight time of the drone under fixed power consumption; t c How long it takes to charge the drone to completion; is the total number of visits to the road segment (i, j) by UAV k in the actual path planning results, 0 means that UAV k has not visited the (i, j) section in the global monitoring task; is the flight distance of UAV k visiting road section (i, j); Δh i and Δv i are the altitude adjustment value of the UAV at intersection i and the corresponding altitude adjustment speed of ascent or descent; t a and t p are the number of monitoring times and the time of hovering and taking photos on the road sections with monitoring tasks in the actual path planning results of the UAV; h a is the flight altitude of the drone on the road section a with monitoring requirements; va represents the visual angle of the drone; d a is the length of section a; And the two objective functions include the following constraints: 1) UAV monitoring accuracy constraint related to flight altitude: f(h)>Acc Where f(h) is the functional relationship between the UAV flight altitude and monitoring accuracy, which needs to be obtained by combining the data analysis and fitting of real scene collection; where Acc is the accuracy requirement, which is a constant; 2) The difference in take-off and landing heights between two adjacent sections of the road to be monitored: Δh a =h a -h a-1 ,a,a-1∈A 3) Drone endurance constraints: in is the remaining flight time of the kth UAV after the monitoring section a ends; Same as d a Consistent, also refers to the length of section a; It is the ratio between the difference between the monitoring height of the previous section and the monitoring height of the next section and the speed of ascent or descent, and the speed of ascent or descent is determined by the positive or negative height difference; Δv is the descent speed to the ground before the battery is exhausted; 4) The road section is monitored at a certain time interval: in is the collection time interval constraint of the monitored road section a, is the time when road section a is monitored for the nth time, is the time when road section a is monitored for the n+1th time, T is the duration of the monitoring task in the entire time period, t a It is the number of monitoring times of the road section a with monitoring tasks in the actual path planning result of the UAV.

2. The multi-UAV road network traffic flow monitoring path optimization method according to claim 1, characterized in that: The multi-objective optimization models all include the following assumptions: The road network for multi-UAV traffic flow monitoring is fixed according to the problem requirements; the monitored time intervals of some sections to be monitored are allowed to be different but always remain unchanged; the power consumption of the UAV during ascent, descent and flight is the same, and only the speed is different; the functional relationship between the UAV flight altitude and the monitoring accuracy is obtained through data analysis and fitting under a large number of different time periods and different environmental conditions, so the monitoring accuracy is no longer affected by the environment; the number and layout of charging stations absolutely meet the requirements for UAVs to charge after performing part of the mission.

3. The multi-UAV road network traffic flow monitoring path optimization method according to claim 2, characterized in that: The road network data includes the coordinates of the starting and ending points of the monitored sections, the connectivity of the non-monitored sections, the intersection coordinates of all sections of the road network, the collection time interval of all sections to be monitored, the flight altitude of the drone in each monitored section, the layout of the charging stations, and the charging time.

4. The multi-UAV road network traffic flow monitoring path optimization method according to claim 1, characterized in that: The step S34 specifically includes: S341, in the process of constructing solutions, the constraint satisfaction is considered, and a population of N feasible solutions is constructed as the initial population; S342, calculating the objective function value: calculating f1 and f2 of each individual in the initial population 1 and the initial population 2; S343, using the f1 and f2 objective function values ​​of all individuals in the population to perform Pareto sorting to obtain individuals with non-dominated objective functions, using the binary tournament selection method to select two excellent parent individuals from the non-dominated individuals to perform crossover operations to generate temporary offspring and perform mutation operations on randomly selected individuals from the temporary offspring, and finally using a local search strategy to guide the offspring individuals to improve their satisfaction of constraints; S344, repeat the above genetic operation until N offspring individuals are generated, and merge the offspring and parent individuals into 2N sets, and then use Pareto non-dominated sorting to retain the N individuals with the best convergence and distribution to form the next generation population; S345, repeat the algorithm process of S341-S344 until the algorithm reaches the maximum number of iterations and terminates; S346, decoding the obtained optimal solution, and exporting and displaying the obtained optimal path set.

5. The multi-UAV road network traffic flow monitoring path optimization method as claimed in claim 4, characterized in that: The crossover operation on the parent individuals in step S343 is specifically as follows: after pairing, the parent individuals randomly selected from the parent population through Pareto sorting and binary tournament have their sequences A and B of equal length in the two individuals exchanged in order from the same corresponding positions.

6. The multi-UAV road network traffic flow monitoring path optimization method according to claim 4, characterized in that: The mutation operation of the temporary offspring individuals in step S343 is specifically as follows: after the crossover operation is performed on the parent individual, an individual is randomly selected from the temporary offspring of the parent population, two numbers are randomly generated as the mutation positions, the information on the two mutation positions of the randomly selected individual is exchanged, and the mutation operation is performed on all other temporary offspring individuals in this way.

7. The multi-UAV road network traffic flow monitoring path optimization method according to claim 4, characterized in that: Step S343 operates the local search strategy after the offspring is generated as follows: after the genetic operations including crossover and mutation are performed on the parent individuals of the population, in order to avoid the genetic operations causing the offspring solution to lose the feasibility of the original parent solution, a local search strategy will be used to change the genes of the offspring individuals to make them more similar to the parent individuals so as to guide the offspring individuals to improve their feasible probability.

Images