Unmanned aerial vehicle nest site selection method based on dream clustering optimization algorithm

Abstract

The unmanned aerial vehicle nest site selection method based on the dream clustering optimization algorithm comprises the following steps: acquiring inspection point data in a forest inspection area; establishing a multi-target mathematical model including the minimum total flight mileage, the minimum routing inspection repeated coverage, the maximum routing inspection coverage range and the shortest routing inspection time; optimizing a k value in the k-means + + cluster by using a sleepwalking optimization algorithm, and generating scattered and reasonable nest site selection points; and global search is carried out by using a dream clustering optimization algorithm, so that the point location of the nest is further optimized. According to the method, the unmanned aerial vehicle nests are arranged in the region, so that the optimal site selection of the unmanned aerial vehicle nests in various regions is realized, the labor cost and resource consumption caused by wrong arrangement of the nests are effectively reduced, and the unmanned aerial vehicles in various regions are satisfied to carry out normalized efficient inspection on a plurality of inspection targets.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) inspection technology, specifically to a UAV nesting site selection method based on the dream clustering optimization algorithm. Background Technology

[0002] With the development of drone technology, drones are widely used in various fields. Forest patrols, due to factors such as terrain, obstacles, and temperature, place higher demands on drone nest location selection. Traditional nest selection methods mostly focus on factors such as geographical location and service coverage, but neglect obstacles in complex environments and the timeliness of patrol tasks. In the complex environment of forests, trees, temperature, and flight altitude significantly affect drone flight paths. Therefore, how to optimize nest location to minimize patrol time remains a problem to be solved.

[0003] In existing technologies, regional inspection has the following problems;

[0004] (1) Complex environmental factors: Climate change is frequent in various regions, and environmental factors such as wind speed, temperature, and precipitation may change suddenly, affecting the flight stability and mission execution of UAVs. The inspection area is large, and there are differences in environmental factors in urban areas, forests, etc., which affect the flight efficiency of UAVs and cause flight missions to fail to be completed normally.

[0005] (2) Flight distance and battery life issues: The battery life of drones is usually limited, especially when patrolling large areas of forest. Long-term flights can easily lead to insufficient battery, affecting the continuity of patrols and the completion of tasks. Some areas may lack sufficient charging infrastructure, which cannot provide timely charging support for drones, resulting in mission interruptions or frequent returns to charging stations, reducing operational efficiency.

[0006] (3) System reliability and safety issues: During long-term flight, the performance of the UAV battery may degrade, leading to battery depletion or even battery failure. System failure may cause the UAV to lose control or crash. Due to the influence of obstacles, wind, and other factors, the UAV may collide with trees or other obstacles during flight, causing damage or being unable to continue its mission. Summary of the Invention

[0007] To address the aforementioned technical issues, this invention discloses a drone nesting location method based on the dream clustering optimization algorithm. By setting up multiple drone nests within a region, the optimal location of drone nests in various regions is achieved, effectively reducing the labor costs and resource consumption caused by incorrect nesting, and meeting the needs of drones in various regions for routine and efficient inspection of multiple inspection targets.

[0008] The technical solution adopted in this invention is as follows:

[0009] The UAV nesting method based on the dream clustering optimization algorithm includes the following steps:

[0010] Step 1: Obtain patrol point data within the forest patrol area;

[0011] Step 2: Establish a multi-objective mathematical model that includes minimum total flight mileage, minimum inspection overlap coverage, maximum inspection coverage area, and shortest inspection time;

[0012] Step 3: Use the sleepwalking optimization algorithm to optimize the k value in k-means++ clustering to generate dispersed and reasonable nesting locations;

[0013] Step 4: Use the dream clustering optimization algorithm to perform a global search, thereby further optimizing the location of the nest.

[0014] In step 1, the set of all inspection points that need to be detected or monitored in the forest inspection area is obtained through GIS remote sensing data or field surveying: N = {1, 2, 3, ..., i}, i ∈ N, i represents a single inspection point, and N represents the set of inspection points;

[0015] Simultaneously, information on the location of obstacles (such as tall trees, hills, and rocks) and flight restriction zones within the forest patrol area is collected to evaluate the feasible flight paths of the UAV in subsequent algorithms; after collecting data from forest patrol points, preprocessing is performed.

[0016] In step 2, establishing a multi-objective mathematical model includes:

[0017] The following four main optimization objectives are selected:

[0018] 1. Minimum total flight distance:

[0019] Let k represent the number of nests, i.e., the number of clusters, and let d ij This represents the flight distance between inspection point i and nest j. Define a binary decision variable:

[0020]

[0021] Where, x ij represents a binary decision variable, where a value of 1 indicates that nest j serves inspection point i, and a value of 0 otherwise.

[0022] The minimum total flight distance can then be expressed as:

[0023]

[0024] Where Z represents the total flight distance of all UAVs during the inspection mission, and the flight distance is shortest when Z is minimum; M represents the cluster of UAVs.

[0025] 2. Minimum inspection overlap coverage area:

[0026] In forest environments, to avoid excessive clustering or repetitive operations of drones in the same area, it is necessary to minimize the overlapping coverage area of ​​the inspection area.

[0027] Let R2 represent the set of inspection points covered by nest j, then the overlapping coverage area is:

[0028]

[0029] Where R1 represents the area repeatedly covered by the inspection, A i B represents the coverage area of ​​nest i. i This represents the area where nest i is repeatedly covered. To make drone inspections more efficient and avoid collisions or interference, R1 should be minimized as much as possible. Whenever there is repeated coverage, R1 will overlap; minimizing the repeated coverage area means minimizing the overlap sum. i Indicates inspection point i;

[0030] 3. Maximum inspection coverage area:

[0031] Considering the large-scale forest patrol mission, the drone's nest should cover as many patrol points as possible geographically to reduce repeated flights and energy consumption caused by omissions; specifically as follows:

[0032]

[0033] Where R2 represents the set of inspection points covered by nest j; maxR2 represents the maximum inspection coverage area of ​​nest j. The maximum inspection coverage area allows the UAV to avoid repeatedly inspecting the same inspection point during inspection tasks, achieving higher inspection efficiency.

[0034] Nest location selection is the foundation of drone inspection. Nests are the basis for the coverage of trees required for forest inspection. Therefore, while ensuring the normal operation of inspection tasks, nest coverage should be maximized as much as possible. At the same time, in order to reduce the risk of drone collisions, the overlap of nest coverage should be minimized and the redundancy of inspecting the same target should be reduced, thereby saving energy and improving inspection efficiency.

[0035] 4. Minimum inspection time:

[0036] To ensure the inspection task is completed within the specified time limit, the flight time objective must be considered. Let T represent the time it takes for the UAV to perform the inspection from its starting point j. Then the overall flight time can be approximately defined as:

[0037]

[0038] Where T represents the total flight time; Z represents the total flight distance of all UAVs during the inspection mission; v represents the UAV inspection flight speed; t ik Indicates the number of inspections at each inspection point; x ik The number of inspections at each inspection point is indicated; t0 represents the drone charging time; t I This indicates the time it takes for the drone to photograph the inspection target during flight; k represents the number of drone nests; minT represents the shortest overall flight time.

[0039] Minimize the flight time, which is the sum of the times of each flight segment, including: fixed inspection flight time, charging time, and the time required to photograph the inspection target. If the flight does not involve charging, j is set to 0. Finally, after all nests and inspection points are covered, minimize T.

[0040] 5. The relevant constraints of the multi-objective mathematical model include:

[0041] 1) To ensure that each inspection point is inspected only once, the constraint on the number of inspections for each inspection point is as follows:

[0042]

[0043] 2) To ensure that the drone remains powered during missions and is fully charged after each recharge from the hive, the following constraints are imposed:

[0044]

[0045] Among them, e' i Let e'0 represent the current battery level of unmanned vehicle i; e'0 represent the battery level at which it leaves the nest after charging is complete; E represents the maximum battery capacity. In step 3, the dream-walking optimization algorithm is used to optimize the k-means++ clustering k value. By simulating the process of subconscious search and awakening correction through the dream clustering optimization algorithm, it is possible to effectively avoid getting trapped in local optima. In the process of optimizing the k value, Levy Flight and Logistic optimization are introduced to enhance the balance between global search capability and local search capability.

[0046] In step 3, the inspection point set i is regarded as the data that needs to be clustered, the number of nests is set to k, and when k-means++ clustering selects the initial cluster center, it will perform weighted random sampling based on the squared distance to the selected cluster center, so that the initial cluster centers are as dispersed as possible. When processing the data, 80% of the data will be retained and 20% of the new data will be added to the initial random sampling data.

[0047] Specifically, it includes the following:

[0048] Step 1. Randomly select the first cluster center:

[0049] A point is randomly selected from the inspection area as the first nest center c1;

[0050] Step 2. Calculate distance and perform weighted extraction:

[0051] For each candidate point i, calculate the distance D(q) from each candidate point to the nearest cluster center among all current cluster centers. l );

[0052]

[0053] Where q represents the candidate point; c represents the cluster center; and C represents the set of cluster centers;

[0054] Then, with D(q) l ) 2 The proportion is used as a probability to select the next cluster center c. n+1 ;

[0055]

[0056] Among them, P i Indicates the new center point; D(i) l ) represents the distance of candidate point i from the cluster center; i l Let i represent candidate point; D(q) represent the distance of candidate point q from the cluster center; q represents the candidate point in the nest; N represents the set of inspection points;

[0057] Step 3. Repeat the iteration until n cluster centers are obtained: forming the initial nest candidate location set {c1, c2, c3, ..., c n}, c1, c2, c3, ..., c n These represent the sets of candidate nest points;

[0058] Step 4. Use k-means++ clustering to iterate quickly through the initial candidate nest points;

[0059] To further refine the first cluster center and make it more closely reflect the data distribution, several iterations are performed on the n cluster centers to obtain a preliminary nesting layout scheme.

[0060] Compared to traditional random initialization, the improved k-means++ clustering provides a more geographically uniform and reasonable nest location, reducing the need for subsequent large-scale "correction" processes. It also takes into account the continuous changes in inspection data and can execute multiple inspection tasks without interruption.

[0061] In step 4, the global search and local search of the Dream Clustering Optimization Algorithm (DCOA) are performed:

[0062] 1) Weighted objective function:

[0063] By weighting four factors—minimum total flight mileage, minimum inspection overlap, maximum inspection coverage, and shortest inspection time—the final optimization target for aircraft nest locations was obtained.

[0064] F(task) = w1Z + w2R1 + w3R2 + w4T

[0065] Where F(task) represents the overall optimization objective; Z represents the total flight mileage; R1 represents the overlapping coverage area; R2 represents the inspection coverage area; T represents the inspection time; w1, w2, w3, and w4 all represent weight coefficients, and w1+w2+w3+w4=1.

[0066] The Dream Clustering Optimization Algorithm (DCOA) effectively avoids getting trapped in local optima by simulating the process of subconscious search and awakening correction.

[0067] In this invention, the Dream Clustering Optimization Algorithm (DCOA) is improved to address the multi-objective characteristics of the nest location problem. By combining strategies such as optimal retention, adaptive search step size, and multi-group cooperation, the algorithm can take into account comprehensive requirements such as coverage, flight time, flight safety, energy consumption, and changes in inspection points in a forest environment.

[0068] 2) Population initialization:

[0069] Define a population where each individual corresponds to a complete nest location or nest-inspection point mapping scheme. Details are as follows:

[0070] Individual = [c1,c2,…,cK], where c1,c2,…,cK represent the nest coordinates, and K represents the number of nests;

[0071] To accelerate convergence, the nest location scheme obtained from k-means++ clustering is used as a preprocessing step; the details are as follows:

[0072] The initial nest locations generated in step 3 are taken as elite individuals; C = {c1, c2, ..., cK}, where k represents the number of nests.

[0073] The initial nest locations are used as the initial data input for the dream clustering algorithm to optimize nest locations, and upper and lower limits for the value of k are defined; specifically as follows:

[0074] Kmin≤K≤Kmax;

[0075] Where: Kmin is the minimum number of nests required to cover all inspection points, and Kmax is the maximum number of nests that can be deployed;

[0076] Simultaneously, the initial nesting point is randomly perturbed to generate other individuals, in order to maintain diversity; specifically as follows:

[0077] Apply a random Gaussian perturbation to the initial nest point set C to generate the remaining N-1 individuals: individual k = C + N(0,σ), k = 2, 3, ..., N individuals

[0078] Where N(0,σ) represents a Gaussian distributed random disturbance with a mean of 0 and a standard deviation of σ. σ is the disturbance intensity, taken as 20% of the standard deviation of the inspection point coordinates.

[0079] The clustering performance of the initial nest locations generated by K-means++ for nest location selection was evaluated using the sum of squared errors within the cluster (SSE), as follows:

[0080]

[0081] p i Indicates the coordinates of the inspection point; c j Indicates the coordinates of the nest location; S j SSE represents the set of inspection points for nest j; K represents the total number of nests; the smaller the SSE, the better the clustering effect.

[0082] 3) Iterative update mechanism:

[0083] a. Subconscious search:

[0084] During the iteration process, each individual will search in the vicinity based on the current optimal solution or historical excellent solutions with a certain probability.

[0085]

[0086] in: Represents the nest coordinate vector of individual i in the (t+1)th generation; Let represent the nest coordinate vector of individual i in generation t; α represents the step size scaling factor of Levi's flight, controlling the exploration range, and is fixed at 0.01; λ t Represents a Lévy random number; c best Indicates the optimal nest coordinates;

[0087] λ t+1 =4λ t (1-λ t ),λ0∈(0,1)

[0088] Where: λ t+1 λ represents the Lévy random number of generation t+1; t Let λ represent the Lévy random number of generation t; λ0 represents the initial Lévy random number; the mapping scheme can enhance the randomness of the search direction to avoid getting trapped in local optima.

[0089] b. Awakening Correction:

[0090] When the Dream Clustering Optimization Algorithm detects that certain nest locations can be locally fine-tuned, it corrects the nest coordinates through local search or short-range k-means iteration, as follows:

[0091] Cj' = cj + η*(cj* - cj)

[0092] Where: Cj' represents the corrected nest coordinates; cj represents the original nest coordinates; η is the step size factor; and cj* is the coordinate of nest j in the local optimal solution.

[0093] The overall inspection objectives, including minimizing inspection time, maximizing nest coverage, and minimizing inspection coverage area, are further optimized, and the global search capability is enhanced through mapping methods.

[0094]

[0095] Among them, L i Indicates step size; δ j Let δj represent the random perturbation vector, where δj = 4δj-1(1-δj-1);

[0096]

[0097] Where: β is the scaling factor, β is 0.01; Γ follows a normal standard distribution; μ represents the environmental sensitivity index; c best The optimal location for the nest;

[0098] It can be made to escape the local optimum by local perturbation;

[0099] c: Select the best option to retain:

[0100] Retain several currently optimal or near-optimal nest locations to prevent the optimal solution from losing the currently optimized nest locations during the update process;

[0101] 4) Nest location optimization:

[0102] ①: Based on the total flight distance, nest overlap coverage, nest coverage area, and inspection time in the above mathematical model, calculate the comprehensive fitness of each individual, that is, change the coefficient w of each function in the weighting function, and retain the solution with higher fitness as a reference for subsequent searches.

[0103] ②: Repeat the iteration until the preset termination conditions are met, including the maximum number of iterations, the convergence threshold, or the number of algebras without improvement.

[0104] 5) Use k-means++ clustering to output the optimal nesting location scheme:

[0105] After the k-means++ clustering algorithm terminates, it outputs the solution with the highest fitness, which is the final nesting point. And the corresponding inspection point—the machine nest inspection mode x ij ; Indicates the optimal nest coordinates;

[0106] At this point, the feasibility of the solution can be further verified according to the requirements, such as terrain obstacle avoidance and battery life. If there are infeasible aspects, a local search or small-scale iteration can be performed again after the constraints are modified.

[0107] This invention provides a method for UAV nesting selection based on the dream clustering optimization algorithm, with the following technical effects:

[0108] 1) The method of this invention can generate high-quality initial nest locations: With the help of k-means++ clustering, the initial nest locations have good dispersion, reducing the large inaccuracies caused by complete randomness.

[0109] 2) The method of this invention has strong global search capability: The Dream Clustering Optimization Algorithm (DCOA) effectively avoids the defect of traditional algorithms that are prone to getting trapped in local optima, and is suitable for scenarios with complex forest terrain and many multi-objective constraints.

[0110] 3) The method of the present invention has good adaptability. Different forest sizes, drone endurance and inspection requirements can be adapted to the needs of different forest scenarios by adjusting the population size, number of iterations and target weights.

[0111] 4) The task of the method of the present invention is expandable: In large-scale forest patrol, it significantly improves the energy consumption management, flight safety and patrol timeliness of UAVs, and can expand the patrol task area during the mission, which has high application value. Attached Figure Description

[0112] The present invention will be further described below with reference to the accompanying drawings and examples;

[0113] Figure 1 The results of the inspection path for the Dream Clustering Optimization Algorithm.

[0114] Figure 2 A schematic diagram of the nest locations after optimization by the Dream Clustering Optimization Algorithm.

[0115] Figure 3 Convergence curves of the Dream Clustering optimization algorithm optimized for Logistic mapping and Lévy flight.

[0116] Figure 4 The flowchart for finding the optimal location of the nest using the dream clustering algorithm. Detailed Implementation

[0117] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0118] The UAV nesting method based on the dream clustering optimization algorithm includes the following steps:

[0119] First, in the data collection and preprocessing stage, geographic information data of all inspection points in the forest inspection area are accurately obtained using geographic information system (GIS) technology. Combined with high-resolution satellite remote sensing images, obstacles such as trees, hills, and rocks in the area that may affect the flight of UAVs are comprehensively marked to form a detailed spatial obstacle information database, providing complete and reliable spatial data support for subsequent flight path planning.

[0120] Secondly, a multi-objective mathematical model for optimizing drone nest location is established, which mainly includes the following optimization objectives: First, minimize the total flight mileage during the drone inspection process to ensure that the inspection task is completed economically and efficiently; second, minimize the degree of redundant coverage within the inspection area to reduce resource waste; third, maximize the inspection coverage of each drone nest, increase the number of inspection points that a single drone nest can serve, and reduce the redundancy of the inspection task; and fourth, while ensuring the smooth execution of the inspection task, optimize the completion time of the drone inspection task to further improve the task execution efficiency.

[0121] To efficiently solve the above optimization model, this embodiment designs an improved dream clustering optimization algorithm. The specific implementation steps are as follows:

[0122] In the initial stage, the k-means++ clustering algorithm is used to preliminarily determine a set of reasonably distributed and evenly distributed candidate nest locations to avoid the initial locations being too concentrated.

[0123] Subsequently, the local search capability of the algorithm is enhanced by introducing the Levy flight strategy. When the algorithm gets stuck in a local optimum, the efficiency of local exploration can be improved by random step size perturbation, which can effectively avoid getting stuck in a local extreme value. At the same time, the algorithm also integrates the Logistic chaotic mapping strategy to significantly enhance the global search capability, generate chaotic sequences, and reset elite individuals with a certain probability, thereby improving the search diversity and global optimization performance of the algorithm.

[0124] Furthermore, the method of this invention uses the sum of squared errors within clusters (SSE) as a fitness function to evaluate the optimization effect of nest location points after clustering. The smaller the SSE value, the better the clustering optimization effect.

[0125] After solving the problem using the improved clustering optimization algorithm described above, the optimal nest location is finally determined using the k-means++ clustering algorithm. The optimized nest layout is more scientific and reasonable, significantly expanding the effective coverage area of ​​the nest, thereby improving the flight efficiency of UAV inspection, reducing energy consumption, and significantly enhancing flight safety.

[0126] Experiments show that the Dream Clustering optimization algorithm can significantly improve the efficiency of forest patrol tasks. The optimized nesting scheme, while ensuring patrol coverage, reduces redundancy in patrol paths and flight time, optimizes UAV energy consumption, and improves mission reliability. Compared to traditional meta-initiative algorithms, the Dream Clustering optimization algorithm, through Logistic mapping and Lévy flight optimization of search capabilities, achieves faster convergence speed. Figure 3 As shown. Figure 3 The convergence performance of the Dream Clustering optimization algorithm and the traditional meta-heuristic algorithm were compared. The horizontal axis represents the number of iterations, and the vertical axis represents the weighted objective function value. The Dream Clustering curve shows a rapid and smooth downward trend, approaching the optimal solution within 100 iterations; the traditional algorithm converges slowly and with significant fluctuations. The advantage stems from the Lévy flight strategy and Logistic chaotic mapping, which reduces the objective function value of Dream Clustering by 25%-40% with the same number of iterations, verifying its ability to efficiently solve multi-objective optimization problems in complex forest scenarios.

[0127] Furthermore, with the same iteration results, the Dream Clustering Optimization Algorithm has a shorter inspection path, such as... Figure 1 As shown. Figure 1 The results of UAV inspection path optimization based on the dream clustering optimization algorithm are presented. The figure clearly marks the distribution of inspection points within the forest area (scattered dots represent targets to be inspected) and the optimized flight path, while also marking the location of drone nests (▲ symbol). Compared with traditional methods, this algorithm significantly shortens the total flight distance, avoids path intersections and redundant coverage; the distribution of drone nests is more balanced, effectively expanding the coverage area and reducing blind spots.

Claims

1. A method for UAV nest selection based on dream clustering optimization algorithm, characterized in that... Includes the following steps: Step 1: Obtain patrol point data within the forest patrol area; Step 2: Establish a multi-objective mathematical model that includes minimum total flight mileage, minimum inspection overlap coverage, maximum inspection coverage area, and shortest inspection time; Step 3: Use the sleepwalking optimization algorithm to optimize the k value in k-means++ clustering to generate dispersed and reasonable nesting locations; Step 4: Use the dream clustering optimization algorithm to perform a global search, thereby further optimizing the location of the nest.

2. The UAV nesting method based on dream clustering optimization algorithm according to claim 1, characterized in that: In step 1, the set of all inspection points that need to be detected or monitored in the forest inspection area is obtained through GIS remote sensing data or field surveying: N = {1, 2, 3, ..., i}, i ∈ N, i represents a single inspection point, and N represents the set of inspection points; Simultaneously, information on the location of obstacles and restricted flight zones within the forest patrol area is collected, and the data from the forest patrol points is preprocessed.

3. The UAV nesting method based on dream clustering optimization algorithm according to claim 1, characterized in that: In step 2, establishing a multi-objective mathematical model includes: Select the following four optimization objectives: 1) Minimum total flight distance: Let k represent the number of nests, i.e., the number of clusters, and let d ij The flight distance between inspection point i and nest j is represented; a binary decision variable is defined: Where, x ij represents a binary decision variable, where a value of 1 indicates that nest j serves inspection point i, and a value of 0 otherwise. The minimum total flight distance is then expressed as: Where Z represents the total flight distance of all UAVs during the inspection mission, and the flight distance is shortest when Z is minimum; M represents the cluster of UAVs. 2) Minimum inspection overlap coverage area: In forest environments, to avoid excessive clustering or repetitive operations of drones in the same area, it is necessary to minimize the overlapping coverage area of ​​the inspection area. Let R2 represent the set of inspection points covered by nest j, then the overlapping coverage area is: Where R1 represents the area repeatedly covered by the inspection, A i B represents the coverage area of ​​nest i. i This represents the area where nest i is repeatedly covered; to make drone inspections more efficient and avoid collisions or interference, R1 should be minimized as much as possible. Whenever there is repeated coverage, R1 will be superimposed; minimizing the repeated coverage area means minimizing the superposition sum; n i Indicates inspection point i; 3) Maximum inspection coverage area: Considering the large-scale forest patrol missions, the drone's nest can cover more patrol points geographically, as shown below: Where R2 represents the set of inspection points covered by nest j; maxR2 represents the maximum inspection coverage of nest j; 4) Minimum inspection time: Let T represent the time it takes for the UAV to perform its inspection from nest j; then the overall flight time is approximately defined as: Where T represents the total flight time; Z represents the total flight distance of all UAVs during the inspection mission; v represents the UAV inspection flight speed; t ik Indicates the number of inspections at each inspection point; x ik The number of inspections at each inspection point is indicated; t0 represents the drone charging time; t I This indicates the time it takes for the drone to photograph the inspection target during flight; k represents the number of drone nests; minT represents the shortest overall flight time. Minimize the flight time, which is the sum of the times of each flight segment, including the fixed inspection flight time, charging time, and the time required to photograph the inspection target. If the flight does not involve charging, j is set to 0. Finally, after all nests and inspection points are covered, minimize T.

4. The UAV nesting method based on dream clustering optimization algorithm according to claim 3, characterized in that: The relevant constraints of the multi-objective mathematical model include: ①: To ensure that each inspection point is inspected only once, the constraint on the number of inspections for each inspection point is as follows: ②: To ensure that the drone remains powered during missions and is fully charged after each recharge from the hive, the following constraints are imposed: Among them, e i ' represents the current battery level of the unmanned vehicle i; e'0 represents the battery level at which the vehicle leaves the nest after charging is complete; E represents the maximum battery capacity.

5. The UAV nesting method based on the dream clustering optimization algorithm according to claim 4, characterized in that: In step 3, the dream-walking optimization algorithm is used to optimize the k-means++ clustering k value; by simulating the process of subconscious search and awakening correction through the dream clustering optimization algorithm, it is possible to effectively avoid getting trapped in local optima; in the process of optimizing the k value, Levy flight and Logistic optimization are introduced to enhance the balance between global search ability and local search ability.

6. The UAV nesting method based on the dream clustering optimization algorithm according to claim 5, characterized in that: Treating the inspection point set i as the data to be clustered, and setting the number of nests as k, k-means++ clustering will perform weighted random sampling based on the squared distance to the selected cluster centers when selecting the initial cluster centers, so that the initial cluster centers are as dispersed as possible. When processing the data, 80% of the data will be retained and 20% of the new data will be added to the initial random sampling data.

7. The UAV nesting method based on the dream clustering optimization algorithm according to claim 6, characterized in that: Step 3 includes the following steps: Step 1. Randomly select the first cluster center: A point is randomly selected from the inspection area as the first nest center c1; Step 2. Calculate distance and perform weighted sampling: For each candidate point i, calculate the distance D(q) from each candidate point to the nearest cluster center among all current cluster centers. l ); Where q represents the candidate point; c represents the cluster center; and C represents the set of cluster centers; Then, with D(q) l ) 2 The proportion is used as a probability to select the next cluster center c. n+1 ; Among them, P i D(i) represents the new center point; l ) represents the distance of candidate point i from the cluster center; i l Let i represent candidate point; D(q) represent the distance of candidate point q from the cluster center; q represents the candidate point in the nest; N represents the set of inspection points; Step 3. Repeat the iteration until n cluster centers are obtained: forming the initial nest candidate location set {c1, c2, c3, ..., c n }, c1, c2, c3, ..., c n These represent the sets of candidate nest points; Step 4. Use k-means++ clustering to iterate quickly on the initial candidate nest points.

8. The UAV nesting method based on the dream clustering optimization algorithm according to claim 7, characterized in that: In step 4, the global search and local search of the dream clustering optimization algorithm are as follows: 1) Weighted objective function: By weighting four factors—minimum total flight mileage, minimum inspection overlap, maximum inspection coverage, and shortest inspection time—the final optimization target for aircraft nest locations was obtained. F(task) = w1Z + w2R1 + w3R2 + w4T Where F(task) represents the overall optimization objective; Z represents the total flight mileage; R1 represents the overlapping coverage area; R2 represents the inspection coverage area; T represents the inspection time; w1, w2, w3, and w4 all represent weight coefficients, and w1+w2+w3+w4=1; 2) Population initialization: Define a population where each individual corresponds to a complete nest location or nest-inspection point mapping scheme; as follows: Individual = [c1,c2,…,cK], where c1,c2,…,cK represent the nest coordinates, and K represents the number of nests; 3) Iterative update mechanism: 4) Nest location optimization: ①: Based on the above mathematical model, the total flight distance, nest overlap coverage area, nest coverage area, and inspection time, calculate the comprehensive fitness of each individual, that is, change the coefficient w of each function in the weighting function, and retain the solution with higher fitness as a reference for subsequent searches; ②: Repeat the iteration until the preset termination conditions are met: including the maximum number of iterations, the convergence threshold, or the number of algebras without improvement; 5) Use k-means++ clustering to output the optimal nesting location scheme: After the k-means++ clustering algorithm terminates, it outputs the solution with the highest fitness, which is the final nesting point. And the corresponding inspection point—the machine nest inspection mode x ij ; Indicates the optimal nest coordinates; At this point, the feasibility of the solution can be further verified according to the requirements. If there are infeasible aspects, a local search or small-scale iteration can be performed again after the constraints are modified.

9. The UAV nesting method based on the dream clustering optimization algorithm according to claim 8, characterized in that: The initial nest locations generated in step 3 are taken as elite individuals; C = {c1, c2, ..., cK}, where k represents the number of nests; The initial nest locations are used as the initial data input for the dream clustering algorithm to optimize nest locations, and upper and lower limits for the value of k are defined; specifically as follows: Kmin≤K≤Kmax; Where: Kmin is the minimum number of nests required to cover all inspection points, and Kmax is the maximum number of nests that can be deployed; Simultaneously, the initial nesting point is randomly perturbed to generate other individuals, in order to maintain diversity; specifically as follows: Apply a random Gaussian perturbation to the initial nest point set C to generate the remaining N-1 individuals: individual k = C + N(0,σ), k = 2, 3, ..., N individuals Where: N(0,σ) is a Gaussian distributed random disturbance with mean 0 and standard deviation σ; σ is the disturbance intensity, which is taken as 20% of the standard deviation of the inspection point coordinates; The clustering performance of the initial nest locations generated by K-means++ for nest location selection was evaluated using the sum of squared errors within the cluster (SSE), as follows: p i Indicates the coordinates of the inspection point; c j Indicates the coordinates of the nest location; S j SSE represents the set of inspection points for nest j; K represents the total number of nests; the smaller the SSE, the better the clustering effect.

10. The UAV nesting method based on the dream clustering optimization algorithm according to claim 8, characterized in that: 3) Iterative update mechanism, specifically including: a. Subconscious search: During the iteration process, each individual will search in the vicinity based on the current optimal solution or historical excellent solutions with a certain probability; in: Represents the nest coordinate vector of individual i in the (t+1)th generation; Let λ represent the nest coordinate vector of individual i in generation t; α represents the step scaling factor of Lévy's flight; t Represents a Lévy random number; c best Indicates the optimal nest coordinates; l t+1 =4min t (1-l t ),λ0∈(0,1) Where: λ t+1 λ represents the Lévy random number of generation t+1; t λ represents the t-th generation Lévy random number; λ0 represents the initial Lévy random number; b is the awakening correction: When the Dream Clustering Optimization Algorithm detects that certain nest locations can be locally fine-tuned, it corrects the nest coordinates through local search or short-range k-means iteration, as follows: Cj' = cj + η*(cj* - cj) Where: Cj' represents the corrected nest coordinates; cj represents the original nest coordinates; η is the step size factor; and cj* is the coordinate of nest j in the local optimal solution. The overall inspection objectives, including minimizing inspection time, maximum nest coverage, and minimizing inspection coverage area, are further optimized, and the global search capability is enhanced through mapping methods. Among them, L i Indicates step size; δ j Represents a random perturbation vector; Where: β is the scaling factor; Γ follows a normal standard distribution; μ represents the environmental sensitivity index, and c best The optimal location for the nest; It can be made to escape the local optimum by local perturbation; c: Select the best option to retain: Retain several currently optimal or near-optimal nest locations to prevent the optimal solution from losing the currently optimized nest locations during the update process.

Images