Power inspection unmanned aerial vehicle path planning method based on fusion of multiple improvement strategies and communication constraints

Abstract

The invention discloses an electric power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion, and the method comprises the steps: (1) scene data collection and customized modeling: employing a cone + local projection composite terrain model in a mountainous area, achieving the adaptive modeling through the height mean value of surrounding houses in an urban area, and carrying out the customized modeling; mountainous area flight height correction and urban area shielding level correction are introduced into channel modeling; (2) improving grey wolf algorithm iterative optimization, and designing a nonlinear convergence factor, a leader wolf historical optimal solution retention mechanism and a wolf pack variation and auxiliary cost fusion strategy; and (3) constructing a five-target track total cost function of obstacle avoidance, routing inspection, length, communication and assistance, and carrying out accumulative punishment on continuous poor communication segments. According to the method, the global optimal solution acquisition rate is improved, and the requirements of high efficiency, safety and data return of power inspection are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle path planning, in particular to a power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion. BACKGROUND

[0002] The power inspection path planning of the unmanned aerial vehicle is a core technology for ensuring the efficiency and safety of the inspection. The existing technology mainly includes two categories: traditional path planning algorithm and bionic intelligent path planning algorithm. However, there are significant limitations: The algorithm optimization ability is insufficient. The time complexity of the Dijkstra algorithm increases with the square of the number of nodes, which is difficult to adapt to large-scale scenarios. The algorithm relies on the heuristic function and is easy to fall into local optimization. The RRT algorithm has high path redundancy and needs additional smoothing optimization. The particle swarm optimization (PSO) is sensitive to the inertia weight and learning factor, and is easy to gather in the later iteration. The genetic algorithm (GA) has weak local search ability. The original grey wolf optimization algorithm (GWO) has linear decay of convergence factor, which is difficult to explore in detail in the later iteration, and the leader wolf does not retain the historical optimal solution, which is easy to appear "iteration regression", and the global optimal solution acquisition rate is low in complex scenarios. The scene modeling precision is low. At present, a single cone is mainly used to simulate the mountain peak, without considering irregular protrusions. For urban houses, OpenStreetMap data is relied on, and the houses without height tags use a fixed default value of 10m. The mountain area directly uses 3GPP RMa scene data without considering the influence of flight height. The ray tracing in the urban area does not distinguish the shielding level of the house, and the channel error rate is high in the high shielding area. At the same time, most of the current schemes only focus on obstacle avoidance and path length, without considering the fusion of communication constraints and path cost function. The planned path is easy to have a communication blind area, which cannot meet the image (such as 1080P) backhaul demand of the power equipment, resulting in the loss or distortion of the inspection data. Therefore, there is an urgent need for a path planning method that takes into account the algorithm optimization ability, scene adaptability and communication reliability to solve the deficiencies of the existing technology. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides a power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion, which solves the problems raised in the background art.

[0004] To achieve the above purpose, the present application is implemented by the following technical scheme: a power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion, comprising the following steps: S1, scene data collection and preprocessing, input satellite images of the inspection area, mountain / house basic data, power inspection point coordinates and channel initial parameters, collect terrain three-dimensional data and real-time channel data through the laser radar and 5G module carried by the unmanned aerial vehicle, and generate the original scene data set; S2, customized modeling, based on the data collected in S1, construct a three-dimensional terrain model and a channel quality model, output an environment model adapted to the inspection scene, wherein the mountainous area adopts a "cone + local protrusion" composite terrain model, the urban area adopts a house height adaptive model, and the mountainous area also introduces a flight height correction and the urban area corrects according to the shielding level; S3, improved grey wolf algorithm iteration optimization, input the environment model of S2 and the preset algorithm parameters, iterate and optimize the path through the nonlinear convergence factor, the leader wolf historical optimal solution reservation mechanism and the wolf swarm mutation strategy, and trigger the total path cost function calculation, output the candidate optimal path; S4, path verification, input the candidate optimal path of S3 and the preset verification standard, judge whether the path is qualified, output the qualified / unqualified result; S5, optimal path output, if S4 verification is passed, output the final optimal path to the unmanned aerial vehicle flight control system; if the verification is not qualified, return to S3 to reiterate and optimize until the qualified optimal path is output.

[0005] Optionally, the "cone + local protrusion" composite terrain model of the mountainous area in step S2 satisfies: ; wherein is the original cone height function, is the height of the i-th mountain peak, is the radius of the bottom surface of the mountain peak, is the Euclidean distance from the point to the center of the bottom surface of the mountain peak, is the protrusion coefficient, and is the total number of mountain peaks in the inspection area. Optionally, the urban house height adaptive model in step S2 satisfies: for a house with a missing height label, the height of the house is calculated based on the height average of the surrounding th house with a height label, that is,

[0006] ; ; wherein , is the calculated height of the house without a height label, is the actual height of the surrounding i-th house with a height label.

[0007] Optionally, the customized channel quality model in step S2 includes: ​​Mountain channel modeling, based on 3GPP RMa scene data, introduce the height of the UAV flight correction term calculation line-of-sight probability , as follows: ; For the height of the UAV flight; Urban channel modeling, based on MATLAB ray tracing toolbox, according to the height and density of the house is divided into three levels of shielding, low level is And Seat / m², medium level is Or Seat / m² (received power multiplied by 0.9-0.95 correction factor), high level is And Seat / m² (received power multiplied by 0.8-0.9 correction factor).

[0008] Optionally, the nonlinear convergence factor in step S3 is in the form of exponential decay or cosine decay; the exponential decay form satisfies , the cosine decay form satisfies ; wherein is the current iteration number, is the maximum iteration number, the value range [0, 2].

[0009] Optionally, the total track cost function in step S3 satisfies the formula as follows: ; Wherein is the collision cost, is the inspection target cost, is the path length cost, is the communication quality cost, is the auxiliary cost.

[0010] Optionally, the wolf swarm mutation strategy in step S3 satisfies: the mutation probability The value is 0.1-0.3, the mountain scene takes 0.2-0.3, the urban scene takes 0.1-0.2; after each iteration, a path node is randomly selected in the probability range, and the coordinates are reinitialized within the UAV flight height range.

[0011] Optionally, the verification standard in step S4 is specifically: ①Collision detection, path node / section and S2 constructed terrain model has no intersection, judged as no collision; ②Inspection coverage, power inspection point and path node Euclidean distance <10-20m, or vertical distance <10-20m and projection in the section, judged as coverage; ③Communication is up to standard, the throughput of all sampling points of the path is greater than or equal to 4-8Mbps, or the number of continuous communication poor sampling points is less than or equal to 2, and it is determined that the communication is up to standard.

[0012] Optionally, the leader wolf historical optimal solution reservation mechanism in step S3 includes two implementation modes: Mode one is direct reservation: define respectively wolf, wolf, wolf's historical optimal position, after each iteration, if the current leader wolf track cost , then , otherwise the historical optimal position is unchanged; Mode two is weighted reservation: update according to the current optimal solution weight 0.6 and the historical optimal solution weight 0.4, that is is the current iteration wolf position.

[0013] Optionally, the communication quality cost in step S3 has two calculation modes, which are continuous punishment and SNR weighting respectively: Continuous punishment: ; ; SNR weighting: ; In the two modes, is the communication cost constant, is the total number of path sampling points.

[0014] The application provides a power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion, which has the following beneficial effects: The multi-strategy improvement of the grey wolf algorithm in the application effectively solves the problems of limited exploration range in the early stage of iteration and difficulty in detailed optimization in the later stage of iteration of the traditional algorithm, and the historical optimal solution reservation mechanism avoids the loss or "backtracking" of the optimal solution in the iteration process, so that the algorithm can more stably explore the global optimal path in a complex inspection scene and is no longer limited by the local optimal solution, thereby providing more reliable algorithm support for path planning. Unlike the traditional simplified modeling mode, the application customizes modeling schemes according to the characteristics of different inspection environments in mountainous areas and urban areas: mountain modeling can restore irregular convex protrusions on the surface of the mountain, urban modeling can fit the actual height difference of houses, channel modeling can adapt to the communication environment characteristics of different scenes, effectively eliminating the error caused by the disconnection between modeling and the actual scene, making the environment model more consistent with the real inspection scene, providing more accurate environmental data basis for path planning, and reducing the flight risk caused by inaccurate modeling. The application integrates communication constraint depth into a track cost function, and designs a special optimization mechanism for a continuous poor communication scene, forces the algorithm to avoid communication blind areas, ensures that the unmanned aerial vehicle can maintain stable communication state during the whole inspection, meets the real-time and clear return demand of key inspection data such as power equipment images, avoids the loss of inspection data caused by communication interruption or poor signal, and guarantees the integrity and effectiveness of the inspection task. The application designs differentiated parameter adaptation logic according to the complexity of the terrain, the density of obstacles and the communication demand difference of different scenes in mountainous areas and urban areas, so that the same technical solution can be flexibly adapted to different types of power inspection environments - it can not only cope with the obstacle avoidance and communication challenges of complex terrain in mountainous areas, but also adapt to the fine path planning demand in the dense housing environment in urban areas, without the need to develop a separate solution for different scenes, significantly improving the versatility and practical value of the technical solution. The multi-target track cost function constructed by the application can realize the comprehensive optimization target of 'no collision, full inspection, short path and optimal communication', the planned path not only avoids obstacles and covers all power inspection points, but also considers path efficiency and communication quality, avoiding the problem of losing one thing to gain another caused by the traditional solution focusing on a single target, so that the safety, integrity, efficiency and reliability of the inspection task are synergistically improved. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The application is a flowchart. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the application will be described clearly and completely in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments.

[0017] In the description of the application, unless otherwise specified, the meaning of 'a plurality of' is two or more; the orientations or positional relationships indicated by the terms 'upper', 'lower', 'left', 'right', 'inner', 'outer', 'front end','rear end', 'head', 'tail' and the like are based on the orientations or positional relationships shown in the drawings, and are only for the purpose of facilitating the description of the application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the application. In addition, the terms 'first','second', 'third' and the like are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0018] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "connected", "connected" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0019] Please refer to Figure 1 The present application provides a technical solution: a power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion, comprising the following steps: S1, scene data acquisition and preprocessing, inputting inspection area satellite image, mountain / house basic data, power inspection point coordinates and channel initial parameters, collecting terrain three-dimensional data and real-time channel data through the laser radar and 5G module carried by the unmanned aerial vehicle, and generating original scene data set; S2, customized modeling, based on the data collected in S1, constructing three-dimensional terrain model (mountainous area adopts "cone + local protrusion" composite terrain model, urban area adopts house height adaptive model) and channel quality model (mountainous area introduces flight height correction, urban area corrects according to shielding level), outputting environment model adapted to inspection scene; The "cone + local protrusion" composite terrain model in mountainous area satisfies: ; Among them is the original cone height function, is the height of the i-th mountain peak, is the radius of the mountain peak bottom surface, is the Euclidean distance from point to the center of the mountain peak bottom surface, is the protrusion coefficient (value 1-3m, high value when the complexity of mountainous terrain is high), is the total number of mountains in the inspection area; The urban house height adaptive model satisfies: for the house with missing height label, the average height of the surrounding houses with height labels is calculated, that is ; Among them (the maximum value is taken when the house density is high), is the calculated height of the house without height label, is the actual height of the surrounding house with height label; The customized channel quality model includes: Mountain channel modeling: Based on 3GPP RMa scene data, introduce the flight height correction term to calculate the line-of-sight probability , as follows: ; for the flight height of the unmanned aerial vehicle; Urban channel modeling: Based on MATLAB ray tracing toolbox, 3 shielding levels are divided according to building height and density, low level is and seat / m² (no correction), medium level is or seat / m² (receive power multiplied by 0.9-0.95 correction factor), high level is and seat / m² (receive power multiplied by 0.8-0.9 correction factor); S3, improve the grey wolf algorithm iteration optimization, input the environment model of S2 and the preset algorithm parameters (maximum iteration number, mutation probability, etc.), optimize the path through the nonlinear convergence factor, leader wolf historical optimal solution reservation mechanism, wolf swarm mutation strategy, trigger the total path cost function calculation (including collision, inspection, length, communication, auxiliary cost) at the same time, output the candidate optimal path; The nonlinear convergence factor is in the form of exponential decay or cosine decay; The exponential decay form satisfies , and the cosine decay form satisfies ; Wherein is the current iteration number, is the maximum iteration number (mountain scene takes 150-200, urban scene takes 100-150), value range [0, 2]; The total path cost function satisfies the formula as follows: ; Wherein is the collision cost, is the inspection target cost, is the path length cost, is the communication quality cost (cumulative punishment is implemented for continuous poor communication section), is the auxiliary cost (including node aggregation, road segment length anomaly, road segment intersection penalty); The wolf swarm mutation strategy satisfies: mutation probability Value 0.1-0.3, mountainous area scene takes 0.2-0.3 (when the terrain is complex, take high value), urban scene takes 0.1-0.2 (when the obstacle is dense, take low value); After each iteration, a path node is randomly selected from the wolf pack (path) in the probability range, and the coordinates are reinitialized within the UAV flight height range (mountainous area 50-80m, urban area 30-60m); The leader wolf historical optimal solution retention mechanism includes two implementation modes: Mode one is direct retention: define Respectively Wolf, Wolf, The historical optimal position of the wolf, after each iteration, if the current leader wolf flight path cost Then , otherwise keep the historical optimal position unchanged; Mode two is weighted retention: update according to the current optimal solution weight 0.6 and the historical optimal solution weight 0.4, that is The current iteration Wolf position; There are two ways to calculate the communication quality cost, which are continuous punishment and SNR weighting: Continuous punishment: ; ; SNR weighting: ; In the two ways, Is the communication cost constant (3000-5000 for mountainous area, 1000-2000 for urban area), Is the total number of path sampling points (according to the UAV flight speed of 15m / s, sampling once every 15m); S4, path verification, input the candidate optimal path of S3 and the preset verification standard (no collision, full inspection coverage, communication up to standard), judge whether the path is qualified or not, output qualified / unqualified result; The verification standard is as follows: ①Collision detection, the path node / section has no intersection with the terrain model constructed in S2, and is determined as no collision; ②Inspection coverage, the power inspection point and the path node Euclidean distance <10-20m, or the vertical distance <10-20m and the projection in the section, and is determined as coverage; ③Communication up to standard, the throughput of all sampling points of the path is ≥4-8Mbps (1080P image backhaul takes 6-8Mbps), or the number of continuous communication poor sampling points is ≤2, and is determined as communication up to standard; S5, optimal path output, if S4 verification passes, output the final optimal path to the UAV flight control system; if the verification is unqualified, return to S3 to reiterate optimization until the qualified optimal path is output.

[0020] Embodiment: Taking DJI M300RTK UAV in mountain water flow protection slope mountain scene as an example, the actual application process of the technical solution of the patent is explained in detail, and the algorithm modules triggered in different stages and the UAV action response are clarified.

[0021] I. Implementation scenario parameter setting (1) Scene environment: The inspection area is a mountain power transmission line section, including 3 water flow protection slope mountains (with irregular protrusions on the surface, slope 35°-45°, vegetation coverage 20%), 2 power tower checkpoints (numbered T1, T2, distance 800m), UAV flight height range 50-80m, communication threshold (Adapt to 1080P power equipment image backhaul); (2) UAV parameters: maximum flight speed 15m / s, turning radius 8m, sampling interval 15m (collect 1 path sampling point every 15m), equipped with 5G communication module and laser radar (for terrain data collection); (3) Algorithm parameters: maximum iteration number , nonlinear convergence factor adopts exponential decay form, wolf swarm mutation probability (high value in mountainous area), collision cost constant , communication cost constant .

[0022] II. Path planning whole process and algorithm triggering logic (1) Scene data acquisition and modeling stage 1. Terrain data acquisition: after the UAV takes off, the water flow protection slope mountain is scanned by the laser radar to obtain the surface three-dimensional coordinate data (including the height and position information of the protruding area), and the coordinates of the power tower checkpoints T1 and T2 are extracted from the satellite image: ; T2: ; 2. Compound terrain modeling: trigger the mountain "cone + local protrusion" modeling algorithm, fit the base shape of the water flow protection slope mountain with a cone with a bottom radius , height , and then correct the irregular protrusions on the surface by the protrusion coefficient to generate an accurate three-dimensional terrain model (collision detection error controlled within 8%); 3. Channel modeling: based on 3GPP RMa scene data, trigger the mountain "height-LOS probability" correlation algorithm, calculate the LOS probability at different positions combined with the current flight height of the UAV (50-80m) , generate a channel quality model (such as the hilly area between T1 and T2, the line-of-sight probability is corrected from 0.6 to 0.75, and the communication quality evaluation deviation is reduced to 6%); (2) Improve the iterative optimization stage of grey wolf algorithm

[0023] 1. Initialization and fitness calculation: the algorithm randomly generates 20 initial paths (each group contains 15-20 path nodes), triggers the total path cost function, and calculates the cost of each path (including collision, inspection, length, communication, and auxiliary 5 types of sub-costs); 2. Nonlinear convergence factor and leader wolf update: early iteration : convergence factor quickly decays from 2 to about 0.8, the algorithm expands the solution space exploration range, avoiding path limitations in a single direction (such as avoiding the steep convex surface of the water flow protection slope mountain, exploring the gentle area of the hilly area); middle iteration : convergence factor , trigger the leader wolf historical optimal solution retention mechanism; if the total cost of the current wolf (optimal path) is lower than the historical optimal cost , update the historical optimal solution ; if the current alpha wolf cost increases in a certain round of iteration , retain the historical optimal solution to avoid "iteration regression"; 3. Wolf swarm mutation and auxiliary cost optimization: after each iteration, trigger the genetic algorithm mutation strategy , randomly select 1 node for each path in 30% of the paths to reinitialize, such as adjusting the node close to the convex from to ; at the same time, trigger the auxiliary cost function to calculate the node aggregation cost If the distance between 2 nodes is 6m < minimum threshold , count and multiply by , road length anomaly , if the length of a certain road segment is 120m > maximum threshold , count the excess length by weight, and the uniformity of the optimized path node distribution improves by 40%, and the road intersection rate decreases to less than 5%; (3) Path verification and optimal path output stage

[0024] 1. Path verification: after 150 iterations of the algorithm, output 3 candidate optimal paths, trigger the path verification algorithm (based on collision detection, inspection coverage, and communication quality 3 standards); for example, candidate path 1 has no collision, covers T1 and T2, but has 2 consecutive communication poor sampling points with throughput of 5.8Mbps ; Candidate path 2 has no collision, full coverage, and 0 poor communication sampling points, with an average path length of 1020m, and is determined to be a qualified path; 2. Optimal path output: Candidate path 2 is output as the final optimal path to the UAV flight control system, with the path nodes being → → →...→ , extending along the gentle area of the water flow protection slope mountain, avoiding protrusions and communication blind areas.

[0025] III. UAV flight execution and algorithm-driven actions

[0026] 1. Path tracking flight: After receiving the optimal path, the UAV flight control system triggers the "node following algorithm" and flies according to the path nodes in sequence; for example, from T1 to the first node , the UAV maintains an altitude of 60m and a speed of 12m / s, and uses laser radar to detect the distance from the mountain in real time. If the distance is less than the safety threshold of 15m, an emergency obstacle avoidance action is triggered, and the heading is adjusted to ; 2. Real-time optimization of communication quality: During flight, the UAV collects 1 communication sampling point every 15m, triggering real-time communication cost calculation; if the throughput of a sampling point drops to 6.5Mbps < 7.8Mbps , and the previous sampling point is also poor , the current , the flight control system adjusts the altitude to 65m according to the algorithm feedback, so that the throughput rises to 7.8Mbps, reset to 0 to avoid continuous poor communication; 3. Checkpoint inspection action: After reaching the power tower T2 , the UAV triggers the "hovering inspection algorithm" based on the inspection target cost in the path planning to confirm the coverage of the checkpoint , then hovers 10m in front of T2, starts the high-definition camera to shoot the power tower components (insulators, wires, etc.), and the image is transmitted back in real time through the 5G module (no communication blind area, with a resolution of 1080P).

[0027] (1) The improved grey wolf algorithm has a planning success rate of 72% (50 rounds of testing), which is 31 percentage points higher than the original grey wolf algorithm (41%), with an average iteration convergence round of 85 rounds, which is 34.6% shorter than the particle swarm algorithm (130 rounds); (2) The collision detection error is reduced to 7.8% in the complex terrain modeling, and there is no collision risk during flight; the average poor communication time is only 4.2s in the communication constraint fusion, which is 90.5% lower than the unconstrained scheme (44.14s); (3) The optimal path length is 1020 m, which is 15.1% shorter than 1201 m planned by the original gray wolf algorithm, and the total time length (including hovering) of the UAV completing T1-T2 inspection is only 8 min, which meets the efficient demand of mountain power inspection; Through MATLAB R2023a simulation, in the mountain (5000x5000x30 m, 13 peaks, 5 checkpoints) and urban (1500x1200x20 m, 80 houses, 5 checkpoints) scenes, the key performance indicators of the improved algorithm and the existing algorithm are compared in Table 1 and Table 2.

[0028] Table 1 Performance comparison in mountain scene (50 rounds of planning)

[0029] Evaluation index Improved algorithm of the patent Original grey wolf algorithm Planning success rate (no collision + all checkpoints) 72% 41% Average poor communication duration (s) 4.2 3.94 Average path length (m) 11200 12010 Average iteration convergence round 85 110 Collision detection error 7.8% 21.5%

[0030] Table 2 Performance comparison in urban scene (50 rounds of planning)

[0031] Evaluation index Improved algorithm of the patent Original grey wolf algorithm Planning success rate (no collision + all checkpoints) 81% 57% Average poor communication duration (s) 2.1 3.94 Average path length (m) 2450 2675.8 Average iteration convergence round 70 95 House modeling error 9.2% 38.5%

[0032] Key parameter values: cost constants such as , According to the scene obstacle density and communication requirements, the obstacle density in the mountain area is low but the avoidance difficulty is high, so the value is higher than that in the urban area; the communication threshold refers to the power industry UAV inspection image transmission standard (such as DL / T1832-2018 "UAV Power Inspection Data Acquisition and Processing Technical Guide"), 1080P image needs to be returned ≥6Mbps; the mutation probability is determined by orthogonal test, and the algorithm convergence speed and optimization ability are optimal in the range of 0.1-0.3.

[0033] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical solution and inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A power inspection unmanned aerial vehicle path planning method based on multi-improvement strategy and communication constraint fusion, characterized in that, Comprise the following steps: S1, scene data acquisition and pretreatment, input inspection area satellite image, mountain / house basic data, power inspection point coordinates and channel initial parameters, through the laser radar and 5G module carried by unmanned aerial vehicle to collect terrain three-dimensional data, real-time channel data, generate original scene data set; S2, customized modeling, based on the data collected in S1, construct three-dimensional terrain model and channel quality model, output the environment model adapted to the inspection scene, wherein the mountainous area adopts "cone+local protrusion" composite terrain model, the urban area adopts house height adaptive model, and the mountainous area also introduces flight height correction and urban area correction according to the shielding level; S3, improved grey wolf algorithm iteration optimization, input the environment model of S2 and preset algorithm parameters, optimize the path through nonlinear convergence factor, leader wolf historical optimal solution reservation mechanism and wolf swarm mutation strategy, and trigger the total path cost function calculation, output the candidate optimal path; S4, path verification, input the candidate optimal path of S3 and the preset verification standard, judge whether the path is qualified, output qualified / unqualified result; S5, optimal path output, if S4 verification passes, output the final optimal path to the unmanned aerial vehicle flight control system; If the verification is not qualified, return to S3 to reiterate optimization until the qualified optimal path is output.

2. The method of claim 1, wherein, The "cone+local protrusion" composite terrain model of the mountainous area in step S2 satisfies: ； wherein is the original cone height function, is the i-th hill height, is the hill base radius, is the Euclidean distance from the point to the hill base center, is the convexity coefficient, is the total number of hills in the inspection area.

3. The method of claim 1, wherein, The adaptive model for building height in urban areas described in step S2 satisfies the following: for buildings lacking height labels, the model adapts to the surrounding environment. The average height of houses with height labels is calculated as follows: ； wherein , is the calculated height for the height-tagged house, is the actual height of the height-tagged house in the neighborhood, is the actual height of the height-tagged house in the neighborhood.

4. The method of claim 1, wherein, The customized channel quality model in step S2 comprises: Mountain channel modeling, based on 3GPP RMa scene data, introduce the height of the UAV flight correction term calculation of line-of-sight probability As follows: ； for the drone flight height; Urban channel modeling, based on MATLAB ray tracing toolbox, 3 shielding levels are divided according to the height and density of houses, the low level is and 0.9-0.95 correction factor) per m2, the medium level is or 0.9-0.95 correction factor) per m2, the high level is and 0.8-0.9 correction factor) per m2.

5. The method of claim 1, wherein, The nonlinear convergence factor in step S3 is in an exponential decay form or a cosine decay form; the exponential decay form satisfies , and the cosine decay form satisfies ; wherein is the current iteration number, is the maximum iteration number, is in the range of [0, 2].

6. The method of claim 1, wherein, The total path cost function in step S3 satisfies the formula as follows: ； wherein is a collision cost, is a patrol target cost, is a path length cost, is a communication quality cost, is an auxiliary cost.

7. The method of claim 1, wherein, The wolf swarm mutation strategy in step S3 satisfies: mutation probability The value is 0.1-0.3, 0.2-0.3 for mountainous area scenes, and 0.1-0.2 for urban area scenes; after each iteration, a path node is randomly selected from the wolf swarm in the probability range, and the coordinates are reinitialized in the flight height range of the unmanned aerial vehicle.

8. The method of claim 1, wherein, The verification standard in step S4 is specifically: ①Collision detection, the path node / section has no intersection with the terrain model constructed in S2, and is determined as no collision; ②Inspection coverage, the power inspection point and the path node Euclidean distance <10-20m, or the vertical distance <10-20m and the projection in the section, and is determined as coverage; ③Communication meets the standard, the throughput of all sampling points of the path is ≥4-8Mbps, or the number of continuous communication poor sampling points is ≤2, and is determined as communication meets the standard.

9. The method of claim 1, wherein, The leader wolf historical optimal solution reservation mechanism in step S3 includes two implementation ways: The first way is direct reservation: definition respectively Wolf, Wolf, The historical optimal position of the wolf, after each iteration, if the current leader wolf path cost Then , otherwise the historical optimal position remains unchanged; The second way is weighted reservation: the current optimal solution weight is 0.6, the historical optimal solution weight is 0.4, and the weighted update is For the current iteration Wolf position.

10. The method of claim 6, wherein, The communication quality cost in step S3 has two calculation ways, which are continuous punishment and SNR weighting: Continuous punishment: ； ； SNR weighting: ； In both ways, is the communication cost constant, is the total number of path samples.

Images