Multi-unmanned aerial vehicle three-dimensional inspection method and system based on improved simulated annealing algorithm

Abstract

The invention provides a multi-unmanned aerial vehicle three-dimensional inspection method and system based on an improved simulated annealing algorithm, and the method comprises the steps: dispatching an unmanned aerial vehicle to collect point cloud data of a region needing inspection, processing the obtained point cloud data, and obtaining the three-dimensional coordinate data of an unmanned aerial vehicle inspection point; according to the intensive degree of the inspection data points and the number of the inspection unmanned aerial vehicles, adopting a K-means + + algorithm to allocate inspection tasks of the unmanned aerial vehicles; and after the routing inspection tasks of the unmanned aerial vehicles are distributed, according to the task routing inspection point three-dimensional coordinates of the unmanned aerial vehicles, an improved simulated annealing algorithm is adopted to optimize routing inspection tracks of all the unmanned aerial vehicles. According to the invention, after the point cloud data is acquired through the unmanned aerial vehicles to obtain the three-dimensional coordinates of the target point needing to be inspected, the inspection task area of each unmanned aerial vehicle is optimized through the clustering method, and then the inspection three-dimensional track of each unmanned aerial vehicle is optimized by adopting the improved simulated annealing algorithm, so that the total inspection time is minimized.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle inspection, and in particular to a multi-unmanned aerial vehicle three-dimensional inspection method and system based on an improved simulated annealing algorithm. Background Art

[0002] With the rapid development of the domestic economy, the mileage of power transmission lines has increased rapidly, and the probability of power line failure has also increased significantly. It is very difficult to locate the fault node from a large range of transmission lines. In particular, due to the long-term exposure of transmission lines to the sun and rain, some key parts are prone to safety hazards, leading to transmission line failures, which will seriously threaten people's lives and social stability. Therefore, it is necessary to conduct accurate, timely and efficient inspections of transmission lines. Traditional inspection methods include manual inspections and helicopter inspections. On the one hand, due to the complexity, wide coverage and dispersion of transmission lines, this brings great difficulties to inspections. On the other hand, the complex terrain and inconvenient transportation in urban or mountainous areas make manual inspections very inconvenient and inefficient, while the cost of helicopter inspections is high and is not conducive to large-scale inspections. This urgently requires us to find a new inspection method to deal with the limitations of existing inspection methods. For this reason, drone inspections came into being. As a modern aviation equipment, drones have played an important role in many areas of life. They not only have the advantages of high mobility, strong timeliness, high probability line-of-sight channels, high positioning accuracy, etc., but also have increasingly low prices, which greatly matches the needs of current inspection tasks. It is worth mentioning that drones have been widely used in power inspections. In terms of inspections, the current inspection trend has gradually shifted from "drone inspections as the main and manual inspections as the auxiliary" to a trend of complete drone inspections.

[0003] Although drones have so many advantages, the limited capacity of drone batteries results in limited endurance and they cannot perform long-term, large-scale inspection tasks. Therefore, how to maximize the use of limited energy for better inspections is one of the keys to improving drone inspection performance.

[0004] Based on this, the present invention aims at the large-scale three-dimensional inspection scenario of multiple UAVs. Under the constraints of the fixed distribution of UAV nests and the power limit of a single UAV inspection, the total inspection time is minimized by optimizing the inspection task allocation of each UAV and the trajectory of the UAV. Summary of the invention

[0005] The present invention provides a multi-UAV three-dimensional inspection method and system based on an improved simulated annealing algorithm. After the UAV collects point cloud data to obtain the three-dimensional coordinates of the target points required for inspection, the inspection task area of ​​each UAV is optimized by a clustering method, and then the improved simulated annealing algorithm is used to optimize the three-dimensional inspection trajectory of each UAV, thereby minimizing the total inspection time.

[0006] A multi-UAV three-dimensional inspection method based on an improved simulated annealing algorithm comprises the following steps:

[0007] Step 101: dispatching a drone to collect point cloud data of the area to be inspected;

[0008] Step 102: Process the acquired point cloud data to obtain the three-dimensional coordinate data of the inspection point of the drone;

[0009] Step 103: After the three-dimensional coordinate data of the inspection points of the UAVs are obtained, the K-means++ algorithm is used to obtain the inspection task area of ​​each UAV according to the density of the inspection data points and the number of inspection UAVs. Assuming that a nest can only store one UAV, the K-means++ algorithm divides the total number of inspection points M into K groups. Assuming that the group center position of the kth group is u k , which contains a total of M inspection points k , the position coordinates of the mth inspection point of the kth group are expressed as α k,m , the purpose of user grouping is to minimize

[0010] Step 104, after the inspection tasks of each drone are assigned, the inspection trajectories of all drones are optimized according to the three-dimensional coordinates of the inspection points of each drone. The improved simulated annealing algorithm combines the 2-optimization method with the traditional simulated annealing algorithm, generates a new trajectory by exchanging two non-continuous edges on the path, and improves the quality of the obtained drone trajectory solution.

[0011] Furthermore, the selection of the user group center in step 103 is proportional to the total distance of the grouped users. The specific steps are as follows:

[0012] 1) Select the initial centroid

[0013] a. Randomly select a patrol point as the first centroid u1=m random ;

[0014] b. For each remaining inspection point m, calculate the distance between them and the nearest centroid in the currently selected centroid Where ||·|| represents the Euclidean distance;

[0015] c. According to the calculated distance, the next centroid is selected probabilistically, where the centroid selection formula is as follows:

[0016]

[0017] That is, the farther the point is, the greater the probability of being selected as the centroid;

[0018] d. Repeat steps b and c until k centroids {u1,u2,...,u k};

[0019] 2) Assign inspection points to the nearest centroid

[0020] a. For each inspection point m, calculate its distance from all centroids u k The Euclidean distance of:

[0021] D(α k,m ,u k )=||α k,m -u k || 2

[0022] b. Assign each inspection point m to the cluster u corresponding to the nearest centroid k , forming k initial clusters;

[0023] 3) Update the centroid

[0024] a. For each cluster, calculate the average coordinates of all inspection points in the cluster and use the average value as the new centroid;

[0025] b. Update the position of the center of mass;

[0026] 4) Iterate steps 2) and 3) until the centroid no longer changes.

[0027] Furthermore, in step 104, an improved simulated annealing algorithm is used to optimize the inspection trajectories of all drones to minimize the sum of the inspection time of all drones, specifically including:

[0028] 1) Set the solution space to q k ={β,1,2,...,M k ,β}, where β is the position of the machine nest;

[0029] 2) Establish the corresponding objective function with the shortest path distance of all inspection points inspected by UAV k as the dependent variable Where D k,i Indicates the trajectory distance of the drone inspection;

[0030] 3) Assume that the solution of the first iteration is Randomly exchange the coordinates of two points u and v and perform 2-opt to generate a new trajectory:

[0031]

[0032] 4) Calculate the difference value ΔE of the objective function, which is expressed as:

[0033] ΔE=f(q k,new )-f(q k,current )

[0034] If ΔE < 0, it means that the trajectory of the new solution is better than the trajectory of the current solution, and the new solution is a better path; if ΔE ≥ 0, then with probability e (-ΔE / T) Set the new solution as the local optimal trajectory, and the solution is selected with probability P * express:

[0035]

[0036] 5) Use the fire reduction coefficient ρ to cool down. When the initial temperature T initial Less than or equal to the termination temperature T final When , output the current local optimal trajectory;

[0037] 6) According to the battery power constraint of the drone, determine whether the energy required for the current trajectory meets the single flight energy constraint of the drone. If so, output the drone inspection trajectory; otherwise, perform binary classification on the drone solution based on the K-means++ algorithm, and then use the simulated annealing algorithm to optimize the trajectory of each classification;

[0038] 7) Finally, check whether there are related clusters that can be merged under the condition of satisfying the single flight constraint of the drone. If so, merge them into the same cluster.

[0039] Furthermore, it also includes:

[0040] Step 105: output the task allocation area of ​​each drone optimized in step 103 and the inspection track of the drone optimized in step 104, and each drone performs the inspection task according to the optimized inspection track.

[0041] A multi-UAV three-dimensional inspection system based on an improved simulated annealing algorithm, comprising:

[0042] Point cloud data collection module, used to collect point cloud data of the area that needs to be inspected;

[0043] The three-dimensional coordinate data acquisition module is used to process the acquired point cloud data and obtain the three-dimensional coordinate data of the drone inspection point;

[0044] The UAV task allocation module is used to obtain the inspection task area of ​​each UAV based on the density of inspection data points and the number of inspection UAVs using the K-means++ algorithm. Assuming that a nest can only store one UAV, the K-means++ algorithm divides the total number of inspection points M into K groups. Assuming that the group center position of the kth group is u k , which contains a total of M inspection points k , the position coordinates of the mth inspection point of the kth group are expressed as α k,m , the purpose of user grouping is to minimize

[0045] The drone trajectory optimization module is used to optimize the inspection trajectories of all drones according to the three-dimensional coordinates of the inspection points of each drone using an improved simulated annealing algorithm. The improved simulated annealing algorithm combines the 2-optimization method with the traditional simulated annealing algorithm, generates a new trajectory by exchanging two non-continuous edges on the path, and improves the quality of the obtained drone trajectory solution.

[0046] Furthermore, when the drone task allocation module selects the user's grouping center, the specific steps are as follows:

[0047] 1) Select the initial centroid

[0048] a. Randomly select a patrol point as the first centroid u1=m random ;

[0049] b. For each remaining inspection point m, calculate the distance between them and the nearest centroid in the currently selected centroid Where ||·|| represents the Euclidean distance;

[0050] c. According to the calculated distance, the next centroid is selected probabilistically, where the centroid selection formula is as follows:

[0051]

[0052] That is, the farther the point is, the greater the probability of being selected as the centroid;

[0053] d. Repeat steps b and c until k centroids {u1,u2,...,u k};

[0054] 2) Assign inspection points to the nearest centroid

[0055] a. For each inspection point m, calculate its distance from all centroids u k The Euclidean distance of:

[0056] D(α k,m ,u k )=||αk,m -u k || 2

[0057] b. Assign each inspection point m to the cluster u corresponding to the nearest centroid k , forming k initial clusters;

[0058] 3) Update the centroid

[0059] a. For each cluster, calculate the average coordinates of all inspection points in the cluster and use the average value as the new centroid;

[0060] b. Update the position of the center of mass;

[0061] 4) Iterate steps 2) and 3) until the centroid no longer changes.

[0062] Furthermore, the UAV trajectory optimization module uses an improved simulated annealing algorithm to optimize the inspection trajectories of all UAVs to minimize the sum of the inspection time of all UAVs, specifically including:

[0063] 1) Set the solution space to q k ={β,1,2,...,M k ,β}, where β is the position of the machine nest;

[0064] 2) Establish the corresponding objective function with the shortest path distance of all inspection points inspected by UAV k as the dependent variable Where D k,i Indicates the trajectory distance of the drone inspection;

[0065] 3) Assume that the solution of the first iteration is Randomly exchange the coordinates of two points u and v and perform 2-opt to generate a new trajectory:

[0066]

[0067] 4) Calculate the difference value ΔE of the objective function, which is expressed as:

[0068] ΔE=f(q k,new )-f(q k,current )

[0069] If ΔE < 0, it means that when the trajectory of the new solution is better than the trajectory of the current solution, the new solution is a better path; if ΔE ≥ 0, with probability e (-ΔE / T) Set the new solution as the local optimal trajectory, and the solution is selected with probability P * express:

[0070]

[0071] 5) Use the fire reduction coefficient ρ to cool down. When the initial temperature T initial Less than or equal to the termination temperature T final When , output the current local optimal trajectory;

[0072] 6) According to the battery power constraint of the drone, determine whether the energy required for the current trajectory meets the single flight energy constraint of the drone. If so, output the drone inspection trajectory; otherwise, perform binary classification on the drone solution based on the K-means++ algorithm, and then use the simulated annealing algorithm to optimize the trajectory of each classification;

[0073] 7) Finally, check whether there are related clusters that can be merged under the condition of satisfying the single flight constraint of the drone. If so, merge them into the same cluster.

[0074] Furthermore, it also includes:

[0075] The result output module is used to output the task allocation area of ​​each drone optimized by the drone task allocation module and the inspection trajectory of the drone optimized by the drone trajectory optimization module. Each drone performs the inspection task according to the optimized inspection trajectory.

[0076] The present invention has the following beneficial effects:

[0077] 1. Use the k-means++ algorithm to assign inspection tasks to each drone. It selects the initial center point in a probability distribution way, making the distance between the initial center points more uniform, which helps to avoid falling into the local optimal solution and make the task allocation result more reasonable;

[0078] 2. Based on the results of inspection task allocation, the two-optimization-simulated annealing algorithm is used to optimize the three-dimensional inspection trajectory of the UAV. Compared with the traditional simulated annealing algorithm, it can not only reduce the total inspection time of the system, but also accelerate the convergence performance of the algorithm;

[0079] 3. Some traditional algorithms are prone to fall into local optimal solutions when searching for the shortest path. The improved two-optimization-simulated annealing algorithm allows a solution that is worse than the current solution to be accepted with a certain probability during the search process, thereby increasing the possibility of obtaining a global optimal solution;

[0080] 4. The present invention has strong scalability. For example, it can solve some inspection problems with the shortest paths or time in tasks such as intelligent logistics distribution, agricultural drone crop monitoring, environmental monitoring, and search and rescue. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required in the embodiments or the prior art description. Obviously, the following drawings and their descriptions are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0082] Figure 1 This is a flow chart of a multi-UAV three-dimensional inspection method based on an improved simulated annealing algorithm according to an embodiment of the present invention;

[0083] Figure 2 This is a flowchart of the UAV inspection task allocation in the embodiment of the present invention.

[0084] Figure 3 The figure is a flowchart of the three-dimensional trajectory design of multiple UAVs in an embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] In order to solve the problem of inspecting a large range of inspection points, the present invention effectively reduces the complexity of the subsequent drone trajectory design by reasonably allocating drone inspection tasks. The inspection trajectory optimization problem is a typical TSP problem. As the number of inspection points increases, the algorithm complexity increases factorially. When faced with a large range of inspection tasks, it is difficult for existing algorithms to obtain effective solutions. Therefore, it is necessary to design an efficient inspection trajectory optimization algorithm to accelerate the convergence of the algorithm. According to the total resources of the system and the limited energy problem, the present invention jointly optimizes the trajectory and task allocation of the drone to minimize the total inspection time consumption of the system.

[0087] See also Figure 1 , an embodiment of the present invention provides a multi-UAV three-dimensional inspection method based on an improved simulated annealing algorithm. Assume that the system includes K UAVs and M transmission line inspection points. These UAVs are located in their respective nests and need to inspect these transmission line inspection points. To perform inspection tasks, the UAVs will take off from the nest and then go to each inspection point to collect key data, such as the status of insulators. However, due to the limited battery capacity of the UAVs, they need to return to the nest for charging before the battery is exhausted. Assuming that the total energy of the UAV's onboard battery is E, the power of the UAV increases linearly with time during the charging process of the nest, that is, E charge=εt, where ε is the charging coefficient, which is a constant. In order to improve the inspection efficiency, it is assumed that the drone flies at a maximum speed V between two inspection points. Every time the drone completes an inspection task and returns to the nest, it must be fully charged before starting the next inspection task. Assuming that the location of the nest and the locations of each inspection point are known, the location of the nest is recorded as β = (x α ,y α ,z α ), the position of the mth transmission line inspection point is α m =(x m ,y m ,z m ).

[0088] Specifically, for UAV k, assume that the total flight time to complete the inspection mission is T k , then the total energy consumption of performing the inspection task is expressed as:

[0089]

[0090] Among them, P(V) represents the instantaneous propulsion energy consumption of the UAV flight process, which is mainly composed of three parts: blade profile power, induced power and parasitic power. The blade profile power is in a cubic relationship with the UAV flight speed, which is mainly used to overcome the resistance of the UAV blades; the induced power is in a square relationship with the UAV flight speed, which is mainly used to overcome the induced resistance of the UAV blades; the parasitic power is in a cubic relationship with the UAV flight speed, which is mainly the power consumed to overcome the fuselage resistance during the flight of the UAV. Specifically expressed as:

[0091]

[0092] Where P0 and P i are two constants, representing the blade profile power and induced power of the drone in hovering state, U tip is the tip speed of the rotor blade, v0 is the average rotor induced speed in the hovering state. In addition, d0 represents the fuselage drag ratio, s represents the rotor solidity, ρ represents the air density, and A represents the rotor disc area.

[0093] The present invention aims to jointly optimize the inspection task allocation and three-dimensional trajectory of each drone under the conditions of the total number of drones and the endurance and airborne energy of each drone, and minimize the total inspection time of the drone group. The total inspection time of drone k includes the inspection time T of the drone flight process k And the drone charging time T k,c . Let the position of drone k at time t be q k,t =(x k,t ,y k,t ,z k,t), the inspection task assigned to drone k is β k , the optimization problem can be modeled as:

[0094]

[0095] stC1:E k ≤E,k=1,2,...,K

[0096] C2:K≤K tot ,

[0097]

[0098] Among them, constraint C1 represents the UAV endurance energy constraint, constraint C2 represents the UAV data quantity constraint, and formula C3 represents the UAV starting point and end point constraints, that is, the UAV takes off from the machine nest and needs to fly to the machine nest for charging after completing a single inspection task.

[0099] The specific steps of the present invention are as follows:

[0100] Step 101: dispatch a drone to collect point cloud data of the area to be inspected. The point cloud data can be obtained by scanning with a drone carrying a 3D laser scanner.

[0101] Step 102: Process the acquired point cloud data to obtain the three-dimensional coordinate data of the drone inspection point.

[0102] Step 103, after the three-dimensional coordinate data of the drone inspection points are obtained, the K-means++ algorithm is used to obtain the inspection task area of ​​each drone according to the density of the inspection data points and the number of inspection drones, assuming that a nest can only store one drone. Compared with the traditional K-means algorithm that randomly selects the initial center, the K-means++ algorithm selects the initial center through a specific probability mechanism, making the algorithm more efficient and able to obtain better clustering results. In addition, the K-means++ algorithm uses distance as an indicator of grouping similarity, so that users with close distances are grouped together. The core idea is to divide the total number of inspection points M into K groups, assuming that the group center position of the kth group is u k , which contains a total of M inspection points k , the position coordinates of the mth inspection point of the kth group are expressed as α k,m , the purpose of user grouping is to minimize The selection of the user group center is random and proportional to the total distance of the group. The specific steps are as follows:

[0103] 1) Select the initial centroid

[0104] a. Randomly select a patrol point as the first centroid u1=mrandom .

[0105] b. For each remaining inspection point m, calculate the distance between them and the nearest centroid in the currently selected centroid where ||·|| represents the Euclidean distance.

[0106] c. According to the calculated distance, the next centroid is selected probabilistically. The centroid selection formula is as follows:

[0107]

[0108] That is, the farther the point is, the greater the probability of being selected as the centroid.

[0109] d. Repeat steps b and c until k centroids {u1,u2,...,u k}.

[0110] 2) Assign inspection points to the nearest centroid

[0111] a. For each inspection point m, calculate its distance from all centroids u k The Euclidean distance of

[0112] D(α k,m ,u k )=||α k,m -u k || 2

[0113] b. Assign each inspection point m to the cluster u corresponding to the nearest centroid k , forming k initial clusters.

[0114] 3) Update the centroid

[0115] a. For each cluster, calculate the average coordinates of all inspection points in the cluster and use the average as the new centroid.

[0116] b. Update the position of the center of mass.

[0117] 4) Iterate steps 2) and 3) until the centroid no longer changes.

[0118] The detailed steps of step 103 are as follows Figure 2 As shown, the specific steps are:

[0119] Step 201: The process starts.

[0120] Step 202: input the extracted three-dimensional inspection point coordinates;

[0121] Step 203: Selecting an initial centroid according to a probability centroid selection formula;

[0122] Step 204: assign inspection points to the nearest centroid;

[0123] Step 205: Update the centroid position;

[0124] Step 206: Determine whether the centroid position has changed, if yes, execute step 207, if no, repeat steps 204-205;

[0125] Step 207: Output the task allocation of each drone.

[0126] Step 104, after the inspection tasks of each drone are assigned, the inspection trajectories of all drones are optimized according to the three-dimensional coordinates of the inspection points of each drone, using an improved simulated annealing algorithm to minimize the sum of the inspection time of all drones. In the process of optimizing the trajectory of the drone, when the drone is low on power, it needs to return to the nest for charging in order to perform subsequent inspection tasks. In the process of optimizing the three-dimensional trajectory of the drone, it is important to minimize the sum of the inspection time of all drones, thereby saving the power of the drone, so that the drone can complete more inspection tasks in a single flight. Specifically, the present invention combines the 2-optimization (2-opt) method with the traditional simulated annealing algorithm, generates a new trajectory by exchanging two non-continuous edges on the path, improves the quality of the obtained drone trajectory solution, and when the path crosses, it means that the current path solution is not optimal. These intersections can be eliminated through 2-opt, thereby generating a shorter path, which helps the simulated annealing algorithm to converge to a trajectory optimization solution close to the optimal one faster.

[0127] The specific steps for optimizing the three-dimensional inspection trajectory of drones to minimize the sum of the inspection time of all drones are as follows:

[0128] 1) Set the solution space to q k ={β,1,2,...,M k ,β}, where β is the position of the machine nest;

[0129] 2) Establish the corresponding objective function with the shortest path distance of all inspection points inspected by UAV k as the dependent variable Where D k,i Indicates the trajectory distance of the drone inspection;

[0130] 3) Assume that the solution of the first iteration is Randomly exchange the coordinates of two points u and v and perform 2-opt to generate a new trajectory:

[0131]

[0132] 4) Calculate the difference value ΔE of the objective function, which is expressed as

[0133] ΔE=f(q k,new )-f(q k,current )

[0134] If ΔE < 0, it means that when the trajectory of the new solution is better than the trajectory of the current solution, the new solution is a better path; if ΔE ≥ 0, with probability e (-ΔE / T) Set the new solution as the local optimal trajectory. The solution can be selected with probability P * express

[0135]

[0136] From the formula P * It can be seen from the expression that the selection of the simulated annealing algorithm solution prevents it from falling into the local optimal solution and is more conducive to finding the optimal solution.

[0137] 5) Use the fire reduction coefficient ρ to cool down. When the initial temperature T initial Less than or equal to the termination temperature T final When , the current local optimal trajectory is output.

[0138] 6) According to the battery power constraint of the UAV, determine whether the energy required for the UAV to complete the current trajectory meets the single flight energy constraint of the UAV. If so, output the inspection trajectory of the UAV; otherwise, perform binary classification on the UAV solution based on the K-means++ algorithm, and then use the simulated annealing algorithm to optimize the trajectory of each classification.

[0139] 7) Finally, check whether there are related clusters that can be merged under the condition of satisfying the single flight constraint of the drone. If so, merge them into the same cluster.

[0140] The detailed steps of step 104 are as follows Figure 3 As shown, the specific steps are:

[0141] Step 301: The process starts.

[0142] Step 302: Set the solution space.

[0143] Step 303: Establish an objective function.

[0144] Step 304: Obtain a new trajectory based on the 2-opt method.

[0145] Step 305: Calculate the difference value of the objective function and select a trajectory solution.

[0146] Step 306: Determine whether the initial temperature is less than the final stability. If so, execute step 307; otherwise, return to step 304.

[0147] Step 307: Determine whether the obtained trajectory solution satisfies the single UAV flight constraint. If so, continue to step 308. Otherwise, divide the clusters that do not meet the constraint into two clusters based on binary classification and return to step 304.

[0148] Step 308: Determine whether there are related clusters that meet the single flight constraint of the drone after merging. If so, merge them into the same cluster and return to step 304. Otherwise, execute step 309.

[0149] Step 309: Output the inspection trajectory of each drone.

[0150] Step 105: output the task allocation area of ​​each drone optimized in step 103 and the inspection track of the drone optimized in step 104, and each drone performs the inspection task according to the optimized inspection track.

[0151] The embodiment of the present invention further provides a multi-UAV three-dimensional inspection system based on an improved simulated annealing algorithm, comprising:

[0152] Point cloud data collection module, used to collect point cloud data of the area that needs to be inspected;

[0153] The three-dimensional coordinate data acquisition module is used to process the acquired point cloud data and obtain the three-dimensional coordinate data of the drone inspection point;

[0154] The UAV task allocation module is used to obtain the inspection task area of ​​each UAV based on the density of inspection data points and the number of inspection UAVs using the K-means++ algorithm. Assuming that a nest can only store one UAV, the K-means++ algorithm divides the total number of inspection points M into K groups. Assuming that the group center position of the kth group is u k , which contains a total of M inspection points k , the position coordinates of the mth inspection point of the kth group are expressed as α k,m , the purpose of user grouping is to minimize

[0155] The drone trajectory optimization module is used to optimize the inspection trajectories of all drones according to the three-dimensional coordinates of the inspection points of each drone using an improved simulated annealing algorithm. The improved simulated annealing algorithm combines the 2-optimization method with the traditional simulated annealing algorithm, generates a new trajectory by exchanging two non-continuous edges on the path, and improves the quality of the obtained drone trajectory solution.

[0156] The result output module is used to output the task allocation area of ​​each drone optimized by the drone task allocation module and the inspection trajectory of the drone optimized by the drone trajectory optimization module. Each drone performs the inspection task according to the optimized inspection trajectory.

[0157] In view of the problem that the limited power of UAVs during inspection makes it difficult to carry out large-scale inspections, and as the number of inspection points increases, the complexity of the UAV trajectory design algorithm increases factorially. When the number of inspection points is large, it is difficult for the existing algorithm to obtain an effective solution. A multi-UAV three-dimensional inspection method and system based on an improved simulated annealing algorithm is proposed, which not only solves the problem that UAVs have limited power and are unable to complete large-scale inspection tasks, but also reduces the complexity of the trajectory algorithm and reduces the inspection time.

[0158] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A multi-UAV three-dimensional inspection method based on an improved simulated annealing algorithm, characterized in that: The steps include: Step 101: dispatching a drone to collect point cloud data of the area to be inspected; Step 102: Process the acquired point cloud data to obtain the three-dimensional coordinate data of the inspection point of the drone; Step 103: After the three-dimensional coordinate data of the inspection points of the UAVs are obtained, the K-means++ algorithm is used to obtain the inspection task area of ​​each UAV according to the density of the inspection data points and the number of inspection UAVs. Assuming that a nest can only store one UAV, the K-means++ algorithm divides the total number of inspection points M into K groups. Assuming that the group center position of the kth group is u k , which contains a total of M inspection points k , the position coordinates of the mth inspection point of the kth group are expressed as α k,m , the purpose of user grouping is to minimize Step 104, after the inspection tasks of each drone are assigned, the inspection trajectories of all drones are optimized according to the three-dimensional coordinates of the inspection points of each drone. The improved simulated annealing algorithm combines the 2-optimization method with the traditional simulated annealing algorithm, generates a new trajectory by exchanging two non-continuous edges on the path, and improves the quality of the obtained drone trajectory solution.

2. The multi-UAV three-dimensional inspection method based on the improved simulated annealing algorithm as claimed in claim 1, characterized in that: The selection of the user group center in step 103 is proportional to the total distance of the grouped users. The specific steps are as follows: 1) Select the initial centroid a. Randomly select a patrol point as the first centroid u1=m random ; b. For each remaining inspection point m, calculate the distance between them and the nearest centroid in the currently selected centroid Where ||·|| represents the Euclidean distance; c. According to the calculated distance, the next centroid is selected probabilistically, where the centroid selection formula is as follows: That is, the farther the point is, the greater the probability of being selected as the centroid; d. Repeat steps b and c until k centroids {u1,u2,...,u k }; 2) Assign inspection points to the nearest centroid a. For each inspection point m, calculate its distance from all centroids u k The Euclidean distance of: D(a k,m ,u k )=||a k,m -u k || 2 b. Assign each inspection point m to the cluster u corresponding to the nearest centroid k , forming k initial clusters; 3) Update the centroid a. For each cluster, calculate the average coordinates of all inspection points in the cluster and use the average value as the new centroid; b. Update the position of the center of mass; 4) Iterate steps 2) and 3) until the centroid no longer changes.

3. The multi-UAV three-dimensional inspection method based on the improved simulated annealing algorithm as claimed in claim 1, characterized in that: In step 104, an improved simulated annealing algorithm is used to optimize the inspection trajectories of all drones to minimize the sum of the inspection time of all drones, specifically including: 1) Set the solution space to q k ={β,1,2,...,M k ,β}, where β is the position of the machine nest; 2) Establish the corresponding objective function with the shortest path distance of all inspection points inspected by UAV k as the dependent variable Where D k,i Indicates the trajectory distance of the drone inspection; 3) Assume that the solution of the first iteration is Randomly exchange the coordinates of two points u and v and perform 2-opt to generate a new trajectory: 4) Calculate the difference value ΔE of the objective function, which is expressed as: ΔE=f(q k,new )-f(q k,current ) If ΔE < 0, it means that the trajectory of the new solution is better than the trajectory of the current solution, and the new solution is a better path; if ΔE ≥ 0, then with probability e (-ΔE / T) Set the new solution as the local optimal trajectory, and the solution is selected with probability P * express: 5) Use the fire reduction coefficient ρ to cool down. When the initial temperature T initial Less than or equal to the termination temperature T final When , output the current local optimal trajectory; 6) According to the battery power constraint of the drone, determine whether the energy required for the current trajectory meets the single flight energy constraint of the drone. If so, output the drone inspection trajectory; otherwise, perform binary classification on the drone solution based on the K-means++ algorithm, and then use the simulated annealing algorithm to optimize the trajectory of each classification; 7) Finally, check whether there are related clusters that can be merged under the condition of satisfying the single flight constraint of the drone. If so, merge them into the same cluster.

4. The multi-UAV three-dimensional inspection method based on the improved simulated annealing algorithm as claimed in claim 1, characterized in that: Also includes: Step 105: output the task allocation area of ​​each drone optimized in step 103 and the inspection track of the drone optimized in step 104, and each drone performs the inspection task according to the optimized inspection track.

5. A multi-UAV three-dimensional inspection system based on an improved simulated annealing algorithm, characterized in that: include: Point cloud data collection module, used to collect point cloud data of the area that needs to be inspected; The three-dimensional coordinate data acquisition module is used to process the acquired point cloud data and obtain the three-dimensional coordinate data of the drone inspection point; The UAV task allocation module is used to obtain the inspection task area of ​​each UAV based on the density of inspection data points and the number of inspection UAVs using the K-means++ algorithm. Assuming that a nest can only store one UAV, the K-means++ algorithm divides the total number of inspection points M into K groups. Assuming that the group center position of the kth group is u k , which contains a total of M inspection points k , the position coordinates of the mth inspection point of the kth group are expressed as α k,m , the purpose of user grouping is to minimize The drone trajectory optimization module is used to optimize the inspection trajectories of all drones according to the three-dimensional coordinates of the inspection points of each drone using an improved simulated annealing algorithm. The improved simulated annealing algorithm combines the 2-optimization method with the traditional simulated annealing algorithm, generates a new trajectory by exchanging two non-continuous edges on the path, and improves the quality of the obtained drone trajectory solution.

6. The multi-UAV three-dimensional inspection system based on the improved simulated annealing algorithm as claimed in claim 5, characterized in that: When the drone task allocation module selects the user's grouping center, the specific steps are as follows: 1) Select the initial centroid a. Randomly select a patrol point as the first centroid u1=m random ; b. For each remaining inspection point m, calculate the distance between them and the nearest centroid in the currently selected centroid Where ||·|| represents the Euclidean distance; c. According to the calculated distance, the next centroid is selected probabilistically, where the centroid selection formula is as follows: That is, the farther the point is, the greater the probability of being selected as the centroid; d. Repeat steps b and c until k centroids {u1,u2,...,u k }; 2) Assign inspection points to the nearest centroid a. For each inspection point m, calculate its distance from all centroids u k The Euclidean distance of: D(a k,m ,u k )=||a k,m -u k || 2 b. Assign each inspection point m to the cluster u corresponding to the nearest centroid k , forming k initial clusters; 3) Update the centroid a. For each cluster, calculate the average coordinates of all inspection points in the cluster and use the average value as the new centroid; b. Update the position of the center of mass; 4) Iterate steps 2) and 3) until the centroid no longer changes.

7. The multi-UAV three-dimensional inspection system based on the improved simulated annealing algorithm as claimed in claim 5, characterized in that: The UAV trajectory optimization module uses an improved simulated annealing algorithm to optimize the inspection trajectories of all UAVs to minimize the sum of the inspection time of all UAVs, specifically including: 1) Set the solution space to q k ={β,1,2,...,M k ,β}, where β is the position of the machine nest; 2) Establish the corresponding objective function with the shortest path distance of all inspection points inspected by UAV k as the dependent variable Where D k,i Indicates the trajectory distance of the drone inspection; 3) Assume that the solution of the first iteration is Randomly exchange the coordinates of two points u and v and perform 2-opt to generate a new trajectory: 4) Calculate the difference value ΔE of the objective function, which is expressed as: ΔE=f(q k,new )-f(q k,current ) If ΔE < 0, it means that the trajectory of the new solution is better than the trajectory of the current solution, and the new solution is a better path; if ΔE ≥ 0, then with probability e (-ΔE / T) Set the new solution as the local optimal trajectory, and the choice of the solution is expressed by probability P*: 5) Use the fire reduction coefficient ρ to cool down. When the initial temperature T initial Less than or equal to the termination temperature T final When , output the current local optimal trajectory; 6) According to the battery power constraint of the drone, determine whether the energy required for the current trajectory meets the single flight energy constraint of the drone. If so, output the drone inspection trajectory; otherwise, perform binary classification on the drone solution based on the K-means++ algorithm, and then use the simulated annealing algorithm to optimize the trajectory of each classification; 7) Finally, check whether there are related clusters that can be merged under the condition of satisfying the single flight constraint of the drone. If so, merge them into the same cluster.

8. The multi-UAV three-dimensional inspection system based on the improved simulated annealing algorithm as claimed in claim 5, characterized in that: Also includes: The result output module is used to output the task allocation area of ​​each drone optimized by the drone task allocation module and the inspection trajectory of the drone optimized by the drone trajectory optimization module. Each drone performs the inspection task according to the optimized inspection trajectory.

Images