UAV inspection path optimization method and system based on mobile UAV nest

Abstract

The present invention relates to the field of patrol path optimization technology, and specifically discloses a method and system for optimizing drone patrol paths based on mobile drone nests. The method includes: performing dynamic environmental perception and multi-source data fusion, generating a global environmental situation map, performing intelligent task analysis and dynamic priority sorting, obtaining a task allocation matrix, performing multi-objective hierarchical path planning, outputting a dynamic flight path set with a timestamp, constructing a real-time obstacle avoidance and emergency self-healing system, obtaining a safe correction path set, performing mobile nest collaborative scheduling and energy optimization, outputting a nest movement instruction set, triggering the update of the global environmental situation map, obtaining drone execution logs, performing knowledge evolution, and completing end-to-end technical iteration. The present invention solves the problem that the route planning of traditional patrol path planning drones fails to achieve refinement and is not highly applicable under complex terrain conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of patrol path optimization, and in particular to a method and system for optimizing the patrol path of an unmanned aerial vehicle (UAV) based on a mobile UAV nest. Background Art

[0002] Mobile drone nests allow drones to be deployed directly to the work site, eliminating the need for manual drone transport, enhancing the drone's emergency response capabilities, and improving operational efficiency. They also provide drones with convenient storage and comprehensive support, such as charging and protection. Furthermore, mobile nests allow for more flexible drone deployment, expanding the scope of inspections and patrols. The continuous advancement of sensor, communication, and artificial intelligence technologies is providing technical support for optimizing drone inspection routes. For example, high-precision lidar sensors enable more accurate terrain mapping and obstacle detection; advanced communication technologies enable real-time data transmission between drones and ground control terminals; and artificial intelligence algorithms can be used for path planning, obstacle avoidance, and other tasks, enhancing drones' autonomous flight capabilities and intelligence.

[0003] Existing drone inspection routes are often manually planned based on on-site terrain conditions and surrounding structures. This approach lacks precision in route planning and is not well-suited for complex terrain. Drones encounter numerous obstacles during flight, and some require manual control to avoid them, which is cumbersome and incapable of automatically avoiding them when collision requirements are met. Furthermore, to maintain a safe distance from obstacles, drones make multiple turns, resulting in turning points and overlapping routes. This results in extended flight distances and high power consumption, impacting inspection efficiency and scope. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a drone patrol path optimization method and system based on a mobile drone nest, which solves the problem that the route planning of traditional patrol path planning drones fails to achieve refinement and is not applicable to complex terrain conditions.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] The method for optimizing the inspection path of UAVs based on a mobile UAV nest includes the following steps:

[0007] Step 1: Perform dynamic environmental perception and multi-source data fusion on the drone inspection area based on the mobile drone nest to generate a global environmental situation map;

[0008] Step 2: Obtain inspection work orders and equipment defect history database, perform intelligent task analysis and dynamic priority sorting based on the global environment situation map, and obtain a task allocation matrix with time windows and priorities;

[0009] Step 3: Collect the drone performance parameters of the mobile drone nest, perform multi-objective hierarchical path planning based on the global environment situation map and the task allocation matrix, and output a dynamic flight path set with a timestamp;

[0010] Step 4: Build a real-time obstacle avoidance and emergency self-healing system based on the dynamic flight path set, obtain a safe correction path set, and conduct drone inspections based on the safe correction path set;

[0011] Step 5: Obtain the remaining power of the mobile nest and road network data, combine them with the safety correction path set, perform mobile nest coordinated scheduling and energy optimization, output the mobile nest movement instruction set, and trigger the update of the global environmental situation map;

[0012] Step 6: Obtain the drone execution log, perform knowledge evolution, extract the characteristics of the drone's successful obstacle avoidance cases, and complete end-to-end technical iteration.

[0013] Preferably, in step 1, the process of generating the global environment situation map is:

[0014] The drone inspection area based on the mobile drone nest performs dynamic environmental perception and multi-source data fusion, including:

[0015] Satellite remote sensing image analysis: Utilizing multispectral analysis technology, the NDVI vegetation index identifies areas where tree growth exceeds the limit. Tree crown height exceeding 5 meters from the vertical distance of the wire is automatically marked. Simultaneously accessing meteorological satellite data, the system obtains cloud distribution and lightning warning information in real time.

[0016] LiDAR point cloud modeling: A ground filtering algorithm is used to separate surface points from tower point clouds. The conductor spatial trajectory is reconstructed using a conductor fitting algorithm, namely the least squares method. The sag error of each span is calculated. If the sag error exceeds the threshold range for each span stored in the database, the segment is marked as an abnormal segment.

[0017] Dynamic obstacle marking: After receiving the civil aviation ADS-B signal, the system analyzes the flight altitude information and generates a three-dimensional cylindrical no-fly zone with a radius of 1 km and an altitude extending to the route definition altitude. The onboard camera uses the YOLOv7 object detection model to identify construction vehicles and dynamically expand the avoidance range.

[0018] Obtain the initial monitoring level of the drone inspection area based on the mobile drone nest;

[0019] Wildfire risk prediction: Integrating historical wildfire data with real-time drought indexes to construct a risk heat map. If the humidity in a region is less than 30% and the wind speed is greater than level 4 for three consecutive days, the regional monitoring level will be automatically raised by one level.

[0020] The risk heat map is created using Geographic Information System (GIS) technology. Historical wildfire data, including location, time, and area, as well as real-time drought indices, including vegetation moisture content and soil moisture, are superimposed on the map. The Kriging interpolation algorithm is used to calculate the risk level of each area, with red indicating high-risk areas and green indicating low-risk areas.

[0021] Machine nest status synchronization: Blockchain technology is used to ensure data consistency among multiple machine nests. When any node updates its location information, other nodes verify the validity of the updated data and add it to the local ledger to achieve status synchronization.

[0022] By integrating satellite remote sensing image analysis, LiDAR point cloud modeling, dynamic obstacle annotation, wildfire risk prediction and machine nest status synchronization, a global environmental situation map is generated.

[0023] Preferably, in step 2, the process of obtaining the task allocation matrix with time windows and priorities is:

[0024] Obtain inspection work orders and perform task semantic analysis based on the global environment situation map:

[0025] Based on the natural language processing engine, the inspection work order text is broken down into: inspection object, target section, and inspection requirements;

[0026] Obtain the priority corresponding to each detection requirement stored in the database;

[0027] Obtain the equipment defect history database, perform dynamic priority adjustments based on the global environment situation map, and perform priority sorting based on the adjusted priorities;

[0028] Resource matching calculation: Based on the drone's camera resolution, flight time, and wind resistance level, a drone capability matrix is ​​constructed. The Euclidean distance algorithm is used to calculate the matching degree with the mission requirements. Combinations with a matching degree of less than 60% are automatically eliminated.

[0029] Get the task assignment matrix with time windows and priorities.

[0030] Preferably, the method of performing dynamic priority adjustment includes:

[0031] Temperature trigger: When the conductor temperature monitoring value in a certain section exceeds the ambient temperature by 30°C and does not drop for 5 minutes, the task upgrade is triggered;

[0032] Compound event processing: If a task involves both lightning strike warning and equipment defect, a dual priority label is generated, marked as urgent + important.

[0033] Preferably, in step 3, the process of outputting the dynamic flight path set with timestamp is:

[0034] Collect drone performance parameters from mobile drone nests, including drone endurance and drone speed;

[0035] Based on the global environment situation map and task allocation matrix, multi-objective hierarchical path planning is carried out, including:

[0036] Global rough planning:

[0037] Genetic algorithm improvements:

[0038] Population initialization: The initial population includes random drone patrol paths, historically optimal drone patrol paths, and manually preset drone patrol paths. In addition to random paths, mutated copies of historically optimal paths are injected, 80% of the core waypoints are retained, and local perturbations are performed on historical paths.

[0039] Crossover strategy: Use sequential crossover to ensure that the tower access order of the child path meets the task priority;

[0040] Dynamic mutation rate: When the optimal solution for five consecutive generations is not improved, the mutation rate is increased to the improved mutation rate stored in the database to escape the local optimum;

[0041] Safety avoidance modeling: A three-dimensional avoidance sphere with a radius of 300 meters is set in the strong electromagnetic zone, and the time cost of path crossing is increased;

[0042] Local scrolling optimization:

[0043] Wind field compensation model: Establishes a UAV aerodynamic model. When the crosswind speed exceeds the crosswind speed threshold stored in the database, the waypoint spacing is automatically increased by 10% to offset the drift error.

[0044] Self-adaptive charging: Calculates the distance covered by the remaining battery in real time. If the remaining battery cannot reach the next nest, a temporary charging task is inserted nearby.

[0045] Airspace deconfliction:

[0046] Altitude layer allocation: emergency missions use the nearest obstacle altitude layer, routine missions use the default altitude layer, and weather monitoring missions use the far-ground altitude layer;

[0047] Intersection negotiation: When paths intersect, the system compares the urgency of the tasks on both sides, and the low-priority drone waits;

[0048] Output a dynamic flight path set with timestamps, including drone waypoint coordinates, drone flight altitude, and drone speed instructions, to ensure that the obstacle avoidance system knows the planned arrival time of each drone.

[0049] Preferably, in step 4, the process of performing drone inspection based on the safety correction path set includes:

[0050] Perception Fusion:

[0051] Millimeter-wave radar: uses frequency-modulated continuous wave mode to analyze the radial velocity of obstacles and predict their positions in the next three seconds;

[0052] Binocular vision: Calculate the obstacle motion vector using dense optical flow method to distinguish between static and dynamic obstacles;

[0053] Obstacle avoidance decision:

[0054] Establish a multi-level early warning mechanism:

[0055] Level 1 warning: When the obstacle distance is greater than the maximum warning distance stored in the database, the path deflection angle is adjusted to maintain the original task progress;

[0056] Level 2 warning: When the closest warning distance stored in the database is less than the obstacle distance and less than the farthest warning distance stored in the database, the Dubins path is activated for quick detour, which increases the time consumption but ensures a safe distance.

[0057] Level 3 emergency: When the obstacle distance is ≤ the nearest warning distance stored in the database, the vehicle is forced to hover and start laser scanning to generate a local high-precision map for re-planning;

[0058] Gradual deviation strategy: When the first level warning is triggered, the camera slowly deviates from the original path at a defined acceleration stored in the database to ensure gimbal shooting stability.

[0059] Dubins path generation: During the second-level warning, a curvature continuous path with the minimum turning radius stored in the database is calculated, and the time increment is controlled within the time increment threshold stored in the database;

[0060] Establishing a group collaboration mechanism;

[0061] Adopt breakpoint resume technology: when the task is interrupted, the completed detection data and breakpoint coordinates are recorded, and the shooting angle is automatically calibrated after restart;

[0062] A safe correction path set is obtained, and the incremental energy consumption data of the detour section is acquired, and drone inspections are carried out based on the safe correction path set.

[0063] Preferably, the establishing of a group collaboration mechanism includes:

[0064] Bidding Agreement: When multiple machines conflict, the system opens a detour priority bidding. For each level of mission urgency, the bidding weight increases by 50%;

[0065] Trajectory negotiation: Exchange path modification plans through V2V communication and select the group avoidance strategy with the shortest overall time.

[0066] Preferably, in step 5, the process of triggering the update of the global environment situation map is:

[0067] Obtain the remaining battery power of the engine and road network data;

[0068] Charging demand forecast:

[0069] Battery health model: Tracks the capacity decay of each charge and discharge cycle of the JiNest battery, obtains the capacity decay rate of each charge and discharge cycle of the JiNest battery, and obtains the remaining capacity of the JiNest battery. Based on the JiNest battery remaining capacity-JiNest charging threshold mapping set in the database, the JiNest charging threshold is dynamically adjusted according to the current JiNest battery remaining capacity;

[0070] Dynamic path planning:

[0071] Docking point calculation: Based on the safety-corrected path set, the estimated arrival time of the drone is obtained, and the time for the drone nest to arrive at the target location is reversely calculated;

[0072] The mobile route of the machine nest prioritizes paved roads and avoids sections with height and width restrictions using OpenStreetMap data;

[0073] Multi-machine nest collaboration:

[0074] Load balancing: When the workload of a certain machine nest exceeds its carrying capacity, it sends a task subcontracting request to three neighboring machine nests. Each neighboring machine nest calculates the feasibility and cost of the subcontracted task based on its current workload and location, and selects the optimal subcontracting solution through an auction negotiation algorithm.

[0075] Output the machine nest movement instruction set, including the machine nest target coordinates, machine nest arrival time, and machine nest path, triggering the update of the global environmental situation map.

[0076] Preferably, in step 6, the process of completing end-to-end technology iteration is:

[0077] Get the drone execution log;

[0078] Feature extraction of successful obstacle avoidance cases performed by the drone: The successful obstacle avoidance trajectory is abstracted into a three-dimensional feature vector consisting of the drone's average speed, the maximum curvature of the obstacle avoidance path, and the total length of the obstacle avoidance path, and stored in the graph database Neo4j;

[0079] Generate a case feature set based on the features of successful obstacle avoidance cases performed by the drone;

[0080] Encrypt the generated case feature set;

[0081] Send the encrypted case feature set to all nests and drones via wireless communication network;

[0082] After receiving the case feature set, the machine nest and the drone perform decryption processing;

[0083] The machine nest and drone are remotely updated based on the decrypted case feature set, and the updated results are fed back to the system to complete end-to-end technical iteration.

[0084] The UAV inspection path optimization system based on the mobile UAV nest is used to implement the above method, including:

[0085] The global environmental situation map generation module is used to perform dynamic environmental perception and multi-source data fusion on the drone patrol area based on the mobile drone nest to generate a global environmental situation map;

[0086] The task allocation matrix acquisition module is used to obtain inspection work orders and equipment defect history databases. Based on the global environment situation map, it performs intelligent task analysis and dynamic priority sorting to obtain a task allocation matrix with time windows and priorities.

[0087] The dynamic flight path set output module is used to collect the performance parameters of the mobile drone nest, perform multi-objective hierarchical path planning based on the global environment situation map and task allocation matrix, and output a dynamic flight path set with a time stamp;

[0088] The emergency self-healing system construction module is used to build a real-time obstacle avoidance and emergency self-healing system based on the dynamic flight path set, obtain a safe correction path set, and conduct drone inspections based on the safe correction path set;

[0089] The global environment situation map update trigger module is used to obtain the remaining power of the machine nest and road network data, combine it with the safety correction path set, perform mobile machine nest coordinated scheduling and energy optimization, output the machine nest movement instruction set, and trigger the global environment situation map update;

[0090] The knowledge evolution module is used to obtain drone execution logs, perform knowledge evolution, extract features of successful obstacle avoidance cases performed by drones, and complete end-to-end technology iteration.

[0091] The present invention has the following beneficial effects:

[0092] This invention generates a global environmental situation map by dynamically sensing the inspection area and fusing multi-source data. This allows the system to fully understand the terrain, obstacle distribution, weather conditions, and other information in the inspection area, providing an accurate basis for subsequent mission planning and path decision-making, and avoiding flight accidents or mission failures caused by unclear environmental information. Combining inspection work orders with a historical database of equipment defects, intelligent task analysis and dynamic priority sorting help to rationally arrange drone inspection tasks based on actual conditions, prioritize important and urgent tasks, improve resource utilization efficiency, and ensure that key areas and equipment are inspected and maintained in a timely manner.

[0093] The present invention performs multi-objective hierarchical path planning based on a global environmental situation map and a task allocation matrix, comprehensively considering task objectives, environmental factors, and UAV performance, to plan a more efficient and reasonable flight path for the UAV, thereby reducing flight time and energy consumption, and improving inspection efficiency and coverage. A real-time obstacle avoidance and emergency self-healing system is constructed based on a dynamic flight path set, so that the UAV can promptly sense and avoid obstacles during flight, and automatically adjust the path when encountering emergencies, ensuring flight safety and reducing the risk of damage to the UAV and mission interruption due to collisions or other accidents. The reasonable movement and energy distribution of the mobile UAV nest are achieved to ensure that the nest can provide support for the UAV in the appropriate location, while avoiding energy waste, extending the overall operation time of the UAV, and carrying out knowledge evolution, so that the path optimization method and the performance of the entire system are continuously improved to better adapt to different inspection scenarios and task requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 Schematic diagram of the overall steps of the method of the present invention;

[0095] Figure 2 Schematic diagram of the detailed steps of the method of the present invention;

[0096] Figure 3 Schematic diagram of module connection of the system of the present invention. DETAILED DESCRIPTION

[0097] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0098] Example 1: Figure 1 、 Figure 2 As shown in FIG, the method for optimizing the inspection path of a UAV based on a mobile UAV nest includes the following steps:

[0099] Step 1: Perform dynamic environmental perception and multi-source data fusion on the drone inspection area based on the mobile drone nest to generate a global environmental situation map;

[0100] Step 2: Obtain inspection work orders and equipment defect history database, perform intelligent task analysis and dynamic priority sorting based on the global environment situation map, and obtain a task allocation matrix with time windows and priorities;

[0101] Step 3: Collect the drone performance parameters of the mobile drone nest, perform multi-objective hierarchical path planning based on the global environment situation map and the task allocation matrix, and output a dynamic flight path set with a timestamp;

[0102] Step 4: Build a real-time obstacle avoidance and emergency self-healing system based on the dynamic flight path set, obtain a safe correction path set, and conduct drone inspections based on the safe correction path set;

[0103] Step 5: Obtain the remaining power of the mobile nest and road network data, combine them with the safety correction path set, perform mobile nest coordinated scheduling and energy optimization, output the mobile nest movement instruction set, and trigger the update of the global environmental situation map;

[0104] Step 6: Obtain the drone execution log, perform knowledge evolution, extract the characteristics of the drone's successful obstacle avoidance cases, and complete end-to-end technical iteration.

[0105] In step 1, the process of generating the global environmental situation map is as follows:

[0106] The drone inspection area based on the mobile drone nest performs dynamic environmental perception and multi-source data fusion, including:

[0107] Satellite remote sensing image analysis: Utilizing multispectral analysis technology, the NDVI vegetation index identifies areas where tree growth exceeds the limit. Tree crown height exceeding 5 meters from the vertical distance of the wire is automatically marked. Simultaneously accessing meteorological satellite data, the system obtains cloud distribution and lightning warning information in real time.

[0108] LiDAR point cloud modeling: A ground filtering algorithm is used to separate surface points from tower point clouds. The conductor spatial trajectory is reconstructed using a conductor fitting algorithm, namely the least squares method. The sag error of each span is calculated. If the sag error exceeds the threshold range for each span stored in the database, the segment is marked as an abnormal segment.

[0109] Dynamic obstacle marking: After receiving the civil aviation ADS-B signal, the system analyzes the flight altitude information and generates a three-dimensional cylindrical no-fly zone with a radius of 1 km and an altitude extending to the route definition altitude. The onboard camera uses the YOLOv7 object detection model to identify construction vehicles and dynamically expand the avoidance range.

[0110] Obtain the initial monitoring level of the drone inspection area based on the mobile drone nest;

[0111] Wildfire risk prediction: Integrating historical wildfire data with real-time drought indexes to construct a risk heat map. If the humidity in a region is less than 30% and the wind speed is greater than level 4 for three consecutive days, the regional monitoring level will be automatically raised by one level.

[0112] The risk heat map is created using Geographic Information System (GIS) technology. Historical wildfire data, including location, time, and area, as well as real-time drought indices, including vegetation moisture content and soil moisture, are superimposed on the map. The Kriging interpolation algorithm is used to calculate the risk level of each area, with red indicating high-risk areas and green indicating low-risk areas.

[0113] Machine nest status synchronization: Blockchain technology is used to ensure data consistency among multiple machine nests. When any node updates its location information, other nodes verify the validity of the updated data and add it to the local ledger to achieve status synchronization.

[0114] By integrating satellite remote sensing image analysis, LiDAR point cloud modeling, dynamic obstacle annotation, wildfire risk prediction and machine nest status synchronization, a global environmental situation map is generated.

[0115] Using multispectral analysis technology and the NDVI vegetation index to identify areas where tree growth exceeds the limit, and automatically marking tree crowns 5 meters vertically above the conductor line, it can detect tree hazards that may affect line safety in advance, provide accurate information for line maintenance and drone inspection planning, and avoid threats posed by trees to transmission lines.

[0116] Accessing meteorological satellite data to obtain cloud distribution and lightning warning information allows drones to proactively avoid adverse weather conditions, ensuring flight safety and helping to rationally schedule inspection times and routes based on weather conditions. By using LiDAR point cloud modeling, ground filtering algorithms, and wire fitting algorithms (least squares method), sag errors are calculated and abnormal sections are marked. This allows for precise monitoring of the status of infrastructure such as transmission lines, timely detection of line anomalies, and improved inspection accuracy and relevance.

[0117] By connecting to civil aviation ADS-B signals to generate no-fly zones, on-board cameras combined with YOLOv7 can identify construction vehicles and expand the avoidance range, effectively avoiding conflicts between drones and flights and ground moving obstacles, ensuring drone flight safety and airspace order.

[0118] By integrating historical wildfire data with real-time drought indices to create a risk heat map, the system automatically increases monitoring levels based on humidity, wind speed, and other conditions. This provides early warning of wildfire risks, enabling drones to more effectively patrol high-risk areas, providing strong support for wildfire prevention and response, and reducing losses caused by wildfires. Blockchain technology ensures data consistency across multiple drone nests, enabling timely and accurate synchronization of information such as the location of each nest node. This facilitates collaborative operations across multiple drone nests, improves the overall operational efficiency and reliability of the mobile drone nest system, and ensures the orderly execution of drone inspection missions.

[0119] In step 2, the process of obtaining the task allocation matrix with time windows and priorities is as follows:

[0120] Obtain inspection work orders and perform task semantic analysis based on the global environment situation map:

[0121] Based on the natural language processing engine, the inspection work order text is broken down into: inspection object, target section, and inspection requirements;

[0122] Obtain the priority corresponding to each detection requirement stored in the database;

[0123] Obtain the equipment defect history database, perform dynamic priority adjustments based on the global environment situation map, and perform priority sorting based on the adjusted priorities;

[0124] Resource matching calculation: Based on the drone's camera resolution, flight time, and wind resistance level, a drone capability matrix is ​​constructed. The Euclidean distance algorithm is used to calculate the matching degree with the mission requirements. Combinations with a matching degree of less than 60% are automatically eliminated.

[0125] Get the task assignment matrix with time windows and priorities.

[0126] Methods for dynamic priority adjustment include:

[0127] Temperature trigger: When the conductor temperature monitoring value in a certain section exceeds the ambient temperature by 30°C and does not drop for 5 minutes, the task upgrade is triggered;

[0128] Compound event processing: If a task involves both lightning strike warning and equipment defect, a dual priority label is generated, marked as urgent + important.

[0129] Using the natural language processing engine to break down the inspection work order text into elements such as the inspection object, target section, and inspection requirements, it can accurately understand the task requirements, avoid ambiguity that may arise from manual interpretation of the work order, and provide an accurate basis for subsequent task planning and resource allocation.

[0130] Dynamic priority adjustments are made based on the database priorities of each inspection requirement, along with the equipment defect history database and global environmental situation map. For example, temperature triggers and complex event handling mechanisms can flexibly increase task priorities based on actual conditions, ensuring that high-risk, urgent tasks are prioritized, allocating resources rationally, and improving inspection response speed to critical issues.

[0131] A drone capability matrix is ​​constructed based on factors such as camera resolution, flight time, and wind resistance. Using the Euclidean distance algorithm, the matrix calculates the degree of compatibility with mission requirements and eliminates poorly matched combinations. This ensures that the drones deployed are compatible with the mission requirements, avoids resource waste, and improves inspection efficiency and quality.

[0132] In step 3, the process of outputting a dynamic flight path set with timestamps is as follows:

[0133] Collect drone performance parameters from mobile drone nests, including drone endurance and drone speed;

[0134] Based on the global environment situation map and task allocation matrix, multi-objective hierarchical path planning is carried out, including:

[0135] Global rough planning:

[0136] Genetic algorithm improvements:

[0137] Population initialization: The initial population includes random drone patrol paths, historically optimal drone patrol paths, and manually preset drone patrol paths. In addition to random paths, mutated copies of historically optimal paths are injected, 80% of the core waypoints are retained, and local perturbations are performed on historical paths.

[0138] Crossover strategy: Use sequential crossover to ensure that the tower access order of the child path meets the task priority;

[0139] Dynamic mutation rate: When the optimal solution for five consecutive generations is not improved, the mutation rate is increased to the improved mutation rate stored in the database to escape the local optimum;

[0140] Safety avoidance modeling: A three-dimensional avoidance sphere with a radius of 300 meters is set in the strong electromagnetic zone, and the time cost of path crossing is increased;

[0141] Local scrolling optimization:

[0142] Wind field compensation model: Establishes a UAV aerodynamic model. When the crosswind speed exceeds the crosswind speed threshold stored in the database, the waypoint spacing is automatically increased by 10% to offset the drift error.

[0143] Self-adaptive charging: Calculates the distance covered by the remaining battery in real time. If the remaining battery cannot reach the next nest, a temporary charging task is inserted nearby.

[0144] Airspace deconfliction:

[0145] Altitude layer allocation: emergency missions use the nearest obstacle altitude layer, routine missions use the default altitude layer, and weather monitoring missions use the far-ground altitude layer;

[0146] Intersection negotiation: When paths intersect, the system compares the urgency of the tasks on both sides, and the low-priority drone waits;

[0147] Output a dynamic flight path set with timestamps, including drone waypoint coordinates, drone flight altitude, and drone speed instructions, to ensure that the obstacle avoidance system knows the planned arrival time of each drone.

[0148] Population initialization integrates random, historically optimal, and manually preset paths. Mutated copies of the historically optimal path are injected and locally perturbed to enrich the search space and increase the probability of finding the global optimal solution. A sequential crossover method ensures that the tower access order of the offspring paths aligns with mission priorities. Dynamic mutation rates prevent the algorithm from falling into local optimality, enabling drones to plan more optimal global inspection paths that better meet mission requirements.

[0149] The wind compensation model accounts for crosswinds and automatically adjusts waypoint spacing to offset drift errors, improving the accuracy of the drone's flight path in complex weather conditions. The adaptive battery function plans power consumption in real time, ensuring the drone can successfully complete missions and replenish energy, improving mission reliability.

[0150] Allocation of flight levels by mission type, such as emergency missions, routine missions, and weather monitoring missions, reduces the risk of airspace conflicts. When routes intersect, lower-priority drones wait, ensuring flight order and safety when multiple drones are operating simultaneously.

[0151] The output dynamic flight path set with timestamps contains information such as waypoint coordinates, flight altitude, speed instructions, etc., providing precise guidance for drone flight and allowing the obstacle avoidance system to grasp the planned arrival time, which is conducive to real-time obstacle avoidance and collaborative operations, improving flight safety and mission execution efficiency.

[0152] In step 4, the process of drone inspection based on the safety correction path set includes:

[0153] Perception Fusion:

[0154] Millimeter-wave radar: uses frequency-modulated continuous wave mode to analyze the radial velocity of obstacles and predict their positions in the next three seconds;

[0155] Binocular vision: Calculate the obstacle motion vector using dense optical flow method to distinguish between static and dynamic obstacles;

[0156] Obstacle avoidance decision:

[0157] Establish a multi-level early warning mechanism:

[0158] Level 1 warning: When the obstacle distance is greater than the maximum warning distance stored in the database, the path deflection angle is adjusted to maintain the original task progress;

[0159] Level 2 warning: When the closest warning distance stored in the database is less than the obstacle distance and less than the farthest warning distance stored in the database, the Dubins path is activated for quick detour, which increases the time consumption but ensures a safe distance.

[0160] Level 3 emergency: When the obstacle distance is ≤ the nearest warning distance stored in the database, the vehicle is forced to hover and start laser scanning to generate a local high-precision map for re-planning;

[0161] Gradual deviation strategy: When the first level warning is triggered, the camera slowly deviates from the original path at a defined acceleration stored in the database to ensure gimbal shooting stability.

[0162] Dubins path generation: During the second-level warning, a curvature continuous path with the minimum turning radius stored in the database is calculated, and the time increment is controlled within the time increment threshold stored in the database;

[0163] Establishing a group collaboration mechanism;

[0164] Adopt breakpoint resume technology: when the task is interrupted, the completed detection data and breakpoint coordinates are recorded, and the shooting angle is automatically calibrated after restart;

[0165] A safe correction path set is obtained, and the incremental energy consumption data of the detour section is acquired, and drone inspections are carried out based on the safe correction path set.

[0166] Establish a group collaboration mechanism, including:

[0167] Bidding Agreement: When multiple machines conflict, the system opens a detour priority bidding. For each level of mission urgency, the bidding weight increases by 50%;

[0168] Trajectory negotiation: Exchange path modification plans through V2V communication and select the group avoidance strategy with the shortest overall time.

[0169] Millimeter-wave radar uses frequency-modulated continuous wave mode to analyze the radial velocity of obstacles and predict their positions in the next three seconds. Binocular vision uses dense optical flow to calculate obstacle motion vectors to distinguish between movement and stillness. It can accurately obtain obstacle information, provide early warnings, buy time for obstacle avoidance decisions, and improve the accuracy and timeliness of obstacle avoidance.

[0170] The multi-level warning mechanism triggers different levels of warning and response strategies at different distances. Level 1 fine-tunes the path to maintain mission progress, level 2 uses a Dubins path for rapid detours to ensure a safe distance, and level 3 provides a forced hover and replanning. This ensures drone safety at varying levels of danger while balancing mission execution and keeping risk and time within reasonable limits.

[0171] The progressive offset strategy ensures gimbal shooting stability, and the Dubins path generation controls the time increment, balancing obstacle avoidance and mission execution requirements.

[0172] When multiple drones conflict, the bidding protocol allocates detour priority according to the urgency of the task, and trajectory negotiation selects the shortest avoidance strategy through V2V communication to avoid confusion caused by multiple drones avoiding obstacles at the same time, improve the efficiency of multi-UAV collaborative operations, and ensure the smooth progress of group operations.

[0173] Breakpoint resume technology records data and coordinates when a task is interrupted, and automatically calibrates the shooting angle after restarting to ensure that the task can continue, reduce data loss, and improve the integrity and reliability of task execution.

[0174] In step 5, the process of triggering the update of the global environment situation map is as follows:

[0175] Obtain the remaining battery power of the engine and road network data;

[0176] Charging demand forecast:

[0177] Battery health model: Tracks the capacity decay of each charge and discharge cycle of the JiNest battery, obtains the capacity decay rate of each charge and discharge cycle of the JiNest battery, and obtains the remaining capacity of the JiNest battery. Based on the JiNest battery remaining capacity-JiNest charging threshold mapping set in the database, the JiNest charging threshold is dynamically adjusted according to the current JiNest battery remaining capacity;

[0178] Dynamic path planning:

[0179] Docking point calculation: Based on the safety-corrected path set, the estimated arrival time of the drone is obtained, and the time for the drone nest to arrive at the target location is reversely calculated;

[0180] The mobile route of the machine nest prioritizes paved roads and avoids sections with height and width restrictions using OpenStreetMap data;

[0181] Multi-machine nest collaboration:

[0182] Load balancing: When the workload of a certain machine nest exceeds its carrying capacity, it sends a task subcontracting request to three neighboring machine nests. Each neighboring machine nest calculates the feasibility and cost of the subcontracted task based on its current workload and location, and selects the optimal subcontracting solution through an auction negotiation algorithm.

[0183] Output the machine nest movement instruction set, including the machine nest target coordinates, machine nest arrival time, and machine nest path, triggering the update of the global environmental situation map.

[0184] Based on the results of load balancing, the task allocation of each machine nest is updated. In the global environmental situation map, the task volume, task priority, and task coverage area of ​​each machine nest are displayed in real time to ensure the transparency and real-time nature of task allocation. In combination with the latest meteorological data, obstacle information (such as no-fly zones and construction vehicle information in dynamic obstacle markings), and wildfire risk prediction results, the environmental information in the global environmental situation map is updated. For example, if the wildfire risk level in a certain area changes, the situation map will promptly adjust the color identification of the area (such as from low-risk green to high-risk red). After completing the above data synchronization and update, the global environmental situation map generation module triggers the global refresh mechanism to push the updated situation map information to each module of the system.

[0185] By tracking the capacity decay of the drone nest's charge and discharge cycles through the battery health model and dynamically adjusting the charging threshold, the drone nest's charging needs can be more accurately predicted, avoiding overcharging or insufficient power, extending the drone nest's battery life, ensuring a stable drone nest energy supply, and providing reliable support for continuous drone operations.

[0186] The arrival time of the machine nest at the target location is calculated in reverse based on the estimated arrival time of the drone, and dynamic path planning is performed. Paved roads are given priority and sections with height and width restrictions are avoided, making the movement path of the machine nest more reasonable, reducing movement time and risks, ensuring that the machine nest can arrive at the designated location in time, and improving the efficiency of collaborative operations between the machine nest and the drone.

[0187] When the machine nest is overloaded with tasks, a task subcontracting request is sent to the neighboring machine nests through the load balancing mechanism, and the optimal subcontracting plan is selected using the auction negotiation algorithm to achieve reasonable distribution of tasks among multiple machine nests, avoid excessive tasks on a single machine nest, and improve the task processing capability and resource utilization efficiency of the entire mobile UAV machine nest system.

[0188] The system outputs a set of instructions for moving the drone nest, including the target coordinates, arrival time, and path of the drone nest, and triggers an update of the global environmental situation map, enabling the system to adjust in real time according to the movement of the drone nest, maintain accurate perception of the environment and mission status, and ensure that the drone inspection mission is carried out continuously, efficiently, and stably.

[0189] In step 6, the process of completing end-to-end technical iteration is as follows:

[0190] Get the drone execution log;

[0191] Feature extraction of successful obstacle avoidance cases performed by the drone: The successful obstacle avoidance trajectory is abstracted into a three-dimensional feature vector consisting of the drone's average speed, the maximum curvature of the obstacle avoidance path, and the total length of the obstacle avoidance path, and stored in the graph database Neo4j;

[0192] Generate a case feature set based on the features of successful obstacle avoidance cases performed by the drone;

[0193] Encrypt the generated case feature set;

[0194] Send the encrypted case feature set to all nests and drones via wireless communication network;

[0195] After receiving the case feature set, the machine nest and the drone perform decryption processing;

[0196] The machine nest and drone are remotely updated based on the decrypted case feature set, and the updated results are fed back to the system to complete end-to-end technical iteration.

[0197] The features of successful obstacle avoidance cases are extracted and stored, abstracted into three-dimensional feature vectors and stored in a graph database to form a case feature set. This allows the drone to accumulate successful obstacle avoidance experience, providing a knowledge base for subsequent flights and facilitating the inheritance and reuse of effective obstacle avoidance strategies.

[0198] By encrypting and transmitting case feature sets, the drone and the pod receive and decrypt them, remotely update them, and provide feedback. This allows the entire system to continuously absorb new successful experiences and continuously improve obstacle avoidance algorithms and strategies based on actual implementation cases, achieving end-to-end technical iteration and enhancing the system's overall obstacle avoidance capabilities and intelligence level.

[0199] As the environment and missions change, new successful obstacle avoidance case features are constantly incorporated into the system, enabling drones and nests to better adapt to complex and diverse scenarios, improve their ability to deal with various potential obstacles and emergencies, and ensure flight safety and the stability of mission execution.

[0200] Example 2: Integrate satellite remote sensing image analysis, LiDAR point cloud modeling, dynamic obstacle annotation, wildfire risk prediction, and machine nest status synchronization to generate a global environmental situation map. The specific integration method is as follows:

[0201] Perform spatiotemporal unified processing on the collected multi-source data and convert them into a unified spatial and temporal coordinate system. Convert data in different formats into a unified format, such as converting satellite remote sensing image data into GeoTIFF format and converting LiDAR point cloud data into LAS format.

[0202] Satellite remote sensing imagery and LiDAR point cloud data can be fused according to geographic spatial location, overlaying the two data in the same geographic coordinate system. This leverages the imagery's spectral information and the point cloud's elevation information to generate a richer topographic feature layer. Dynamic obstacle annotations are updated in real time to their corresponding spatial locations based on their dynamic changing characteristics.

[0203] High-precision data is given a higher weight during fusion. For example, when fusing satellite remote sensing imagery and LiDAR point cloud data, if the LiDAR point cloud data is more accurate, then when generating topographic features, the LiDAR data will be used as the primary fusion, supplemented by the satellite remote sensing imagery.

[0204] For data with high real-time requirements, such as dynamic obstacle annotation and wildfire risk prediction results, the latest data is prioritized for fusion and promptly updated to the global environmental situation map. For relatively stable data, such as satellite remote sensing imagery and LiDAR point cloud data, regular updates and fusion can be performed.

[0205] Fusion algorithms, such as Kalman filtering, DS evidence theory, or machine learning-based algorithms, combine information from different data sources to generate a unified environmental situation dataset. This fused data is then subjected to noise reduction and smoothing to remove false targets and abnormal information, highlight key information, and form a dataset that accurately reflects the environmental situation in the patrol area.

[0206] Prepare geographic information data for the inspection area, including basic geographic elements such as topography, river systems, road networks, and settlements, to build a basic geographic framework. Categorize the fused data according to different situational elements, creating multiple layers such as vegetation, power facilities, obstacles, environmental risks, and machine nests. Use GIS software to render and visualize each layer, using different colors, symbols, and line types to visually display the environmental situation of the inspection area and generate a global environmental situation map.

[0207] Example 3: Figure 3 As shown, a drone inspection path optimization system based on a mobile drone nest is used to implement the method in Example 1 or Example 2, including:

[0208] The global environmental situation map generation module is used to perform dynamic environmental perception and multi-source data fusion on the drone patrol area based on the mobile drone nest to generate a global environmental situation map;

[0209] The task allocation matrix acquisition module is used to obtain inspection work orders and equipment defect history databases. Based on the global environment situation map, it performs intelligent task analysis and dynamic priority sorting to obtain a task allocation matrix with time windows and priorities.

[0210] The dynamic flight path set output module is used to collect the performance parameters of the mobile drone nest, perform multi-objective hierarchical path planning based on the global environment situation map and task allocation matrix, and output a dynamic flight path set with a time stamp;

[0211] The emergency self-healing system construction module is used to build a real-time obstacle avoidance and emergency self-healing system based on the dynamic flight path set, obtain a safe correction path set, and conduct drone inspections based on the safe correction path set;

[0212] The global environment situation map update trigger module is used to obtain the remaining power of the machine nest and road network data, combine it with the safety correction path set, perform mobile machine nest coordinated scheduling and energy optimization, output the machine nest movement instruction set, and trigger the global environment situation map update;

[0213] The knowledge evolution module is used to obtain drone execution logs, perform knowledge evolution, extract features of successful obstacle avoidance cases performed by drones, and complete end-to-end technology iteration.

Claims

1. A UAV inspection path optimization method based on a mobile UAV nest is characterized by: The following steps are involved: Step 1: Perform dynamic environmental perception and multi-source data fusion on the drone inspection area based on the mobile drone nest to generate a global environmental situation map; Step 2: Obtain inspection work orders and equipment defect history database, perform intelligent task analysis and dynamic priority sorting based on the global environment situation map, and obtain a task allocation matrix with time windows and priorities; Step 3: Collect the drone performance parameters of the mobile drone nest, perform multi-objective hierarchical path planning based on the global environment situation map and the task allocation matrix, and output a dynamic flight path set with a timestamp; Step 4: Build a real-time obstacle avoidance and emergency self-healing system based on the dynamic flight path set, obtain a safe correction path set, and conduct drone inspections based on the safe correction path set; Step 5: Obtain the remaining power of the mobile nest and road network data, combine them with the safety correction path set, perform mobile nest coordinated scheduling and energy optimization, output the mobile nest movement instruction set, and trigger the update of the global environmental situation map. The process is as follows: Obtain the remaining battery power of the engine and road network data; Charging demand forecast: Battery health model: Tracks the capacity decay of each charge and discharge cycle of the JiNest battery, obtains the capacity decay rate of each charge and discharge cycle of the JiNest battery, and obtains the remaining capacity of the JiNest battery. Based on the JiNest battery remaining capacity-JiNest charging threshold mapping set in the database, the JiNest charging threshold is dynamically adjusted according to the current JiNest battery remaining capacity; Dynamic path planning: Docking point calculation: Based on the safety-corrected path set, the estimated arrival time of the drone is obtained, and the time for the drone nest to arrive at the target location is reversely calculated; The mobile route of the machine nest prioritizes paved roads and avoids sections with height and width restrictions using OpenStreetMap data; Multi-machine nest collaboration: Load balancing: When the workload of a certain machine nest exceeds its carrying capacity, it sends a task subcontracting request to three neighboring machine nests. Each neighboring machine nest calculates the feasibility and cost of the subcontracted task based on its current workload and location, and selects the optimal subcontracting solution through an auction negotiation algorithm. Output the nest movement instruction set, including the nest target coordinates, nest arrival time, and nest path, triggering the update of the global environment situation map; Step 6: Obtain the drone execution log, perform knowledge evolution, extract the characteristics of the drone's successful obstacle avoidance cases, and complete end-to-end technical iteration.

2. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 1, characterized in that: In step 1, the process of generating the global environment situation map is as follows: The drone inspection area based on the mobile drone nest performs dynamic environmental perception and multi-source data fusion, including: Satellite remote sensing image analysis: Utilizing multispectral analysis technology, the NDVI vegetation index identifies areas where tree growth exceeds the limit. Tree crown height exceeding 5 meters from the vertical distance of the wire is automatically marked. Simultaneously accessing meteorological satellite data, the system obtains cloud distribution and lightning warning information in real time. LiDAR point cloud modeling: A ground filtering algorithm is used to separate surface points from tower point clouds. The conductor spatial trajectory is reconstructed using a conductor fitting algorithm, namely the least squares method. The sag error of each span is calculated. If the sag error exceeds the threshold range for each span stored in the database, the segment is marked as an abnormal segment. Dynamic obstacle marking: After receiving the civil aviation ADS-B signal, the system analyzes the flight altitude information and generates a three-dimensional cylindrical no-fly zone with a radius of 1 km and an altitude extending to the route definition altitude. The onboard camera uses the YOLOv7 object detection model to identify construction vehicles and dynamically expand the avoidance range. Obtain the initial monitoring level of the drone inspection area based on the mobile drone nest; Wildfire risk prediction: Integrating historical wildfire data with real-time drought indexes to construct a risk heat map. If the humidity in a region is less than 30% and the wind speed is greater than level 4 for three consecutive days, the regional monitoring level will be automatically raised by one level. The risk heat map is created using Geographic Information System (GIS) technology. Historical wildfire data, including location, time, and area, as well as real-time drought indices, including vegetation moisture content and soil moisture, are superimposed on the map. The Kriging interpolation algorithm is used to calculate the risk level of each area, with red indicating high-risk areas and green indicating low-risk areas. Machine nest status synchronization: Blockchain technology is used to ensure data consistency among multiple machine nests. When any node updates its location information, other nodes verify the validity of the updated data and add it to the local ledger to achieve status synchronization. By integrating satellite remote sensing image analysis, LiDAR point cloud modeling, dynamic obstacle annotation, wildfire risk prediction and machine nest status synchronization, a global environmental situation map is generated.

3. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 1, characterized in that: In step 2, the process of obtaining the task allocation matrix with time windows and priorities is as follows: Obtain inspection work orders and perform task semantic analysis based on the global environment situation map: Based on the natural language processing engine, the inspection work order text is broken down into: inspection object, target section, and inspection requirements; Obtain the priority corresponding to each detection requirement stored in the database; Obtain the equipment defect history database, perform dynamic priority adjustments based on the global environment situation map, and perform priority sorting based on the adjusted priorities; Resource matching calculation: Based on the drone's camera resolution, flight time, and wind resistance level, a drone capability matrix is ​​constructed. The Euclidean distance algorithm is used to calculate the matching degree with the mission requirements. Combinations with a matching degree of less than 60% are automatically eliminated. Get the task assignment matrix with time windows and priorities.

4. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 3 is characterized in that: The method of performing dynamic priority adjustment includes: Temperature trigger: When the conductor temperature monitoring value in a certain section exceeds the ambient temperature by 30°C and does not drop for 5 minutes, the task upgrade is triggered; Compound event processing: If a task involves both lightning strike warning and equipment defect, a dual priority label is generated, marked as urgent + important.

5. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 1, characterized in that: In step 3, the process of outputting a dynamic flight path set with a timestamp is as follows: Collect drone performance parameters from mobile drone nests, including drone endurance and drone speed; Based on the global environment situation map and task allocation matrix, multi-objective hierarchical path planning is carried out, including: Global rough planning: Genetic algorithm improvements: Population initialization: The initial population includes random drone patrol paths, historically optimal drone patrol paths, and manually preset drone patrol paths. In addition to random paths, mutated copies of historically optimal paths are injected, 80% of the core waypoints are retained, and local perturbations are performed on historical paths. Crossover strategy: Use sequential crossover to ensure that the tower access order of the child path meets the task priority; Dynamic mutation rate: When the optimal solution for five consecutive generations is not improved, the mutation rate is increased to the improved mutation rate stored in the database to escape the local optimum; Safety avoidance modeling: A three-dimensional avoidance sphere with a radius of 300 meters is set in the strong electromagnetic zone, and the time cost of path crossing is increased; Local scrolling optimization: Wind field compensation model: Establishes a UAV aerodynamic model. When the crosswind speed exceeds the crosswind speed threshold stored in the database, the waypoint spacing is automatically increased by 10% to offset the drift error. Self-adaptive charging: Calculates the distance covered by the remaining battery in real time. If the remaining battery cannot reach the next nest, a temporary charging task is inserted nearby. Airspace deconfliction: Altitude layer allocation: emergency missions use the nearest obstacle altitude layer, routine missions use the default altitude layer, and weather monitoring missions use the far-ground altitude layer; Intersection negotiation: When paths intersect, the system compares the urgency of the tasks on both sides, and the low-priority drone waits; Output a dynamic flight path set with timestamps, including drone waypoint coordinates, drone flight altitude, and drone speed instructions, to ensure that the obstacle avoidance system knows the planned arrival time of each drone.

6. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 1, characterized in that: In step 4, the process of performing drone inspection based on the safety correction path set includes: Perception Fusion: Millimeter-wave radar: uses frequency-modulated continuous wave mode to analyze the radial velocity of obstacles and predict their positions in the next three seconds; Binocular vision: Calculate the obstacle motion vector using dense optical flow method to distinguish between static and dynamic obstacles; Obstacle avoidance decision: Establish a multi-level early warning mechanism: Level 1 warning: When the obstacle distance is greater than the maximum warning distance stored in the database, the path deflection angle is adjusted to maintain the original task progress; Level 2 warning: When the closest warning distance stored in the database is less than the obstacle distance and less than the farthest warning distance stored in the database, the Dubins path is activated for quick detour, which increases the time consumption but ensures a safe distance. Level 3 emergency: When the obstacle distance is ≤ the nearest warning distance stored in the database, the vehicle is forced to hover and start laser scanning to generate a local high-precision map for re-planning; Gradual deviation strategy: When the first level warning is triggered, the camera slowly deviates from the original path at a defined acceleration stored in the database to ensure gimbal shooting stability. Dubins path generation: During the second-level warning, a curvature continuous path with the minimum turning radius stored in the database is calculated, and the time increment is controlled within the time increment threshold stored in the database; Establishing a group collaboration mechanism; Adopt breakpoint resume technology: when the task is interrupted, the completed detection data and breakpoint coordinates are recorded, and the shooting angle is automatically calibrated after restart; A safe correction path set is obtained, and the incremental energy consumption data of the detour section is acquired, and drone inspections are carried out based on the safe correction path set.

7. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 6, characterized in that: The establishment of a group collaboration mechanism includes: Bidding Agreement: When multiple machines conflict, the system opens a detour priority bidding. For each level of mission urgency, the bidding weight increases by 50%; Trajectory negotiation: Exchange path modification plans through V2V communication and select the group avoidance strategy with the shortest overall time.

8. The method for optimizing the inspection path of a UAV based on a mobile UAV nest according to claim 1, characterized in that: In step 6, the process of completing end-to-end technology iteration is as follows: Get the drone execution log; Feature extraction of successful obstacle avoidance cases performed by the drone: The successful obstacle avoidance trajectory is abstracted into a three-dimensional feature vector consisting of the drone's average speed, the maximum curvature of the obstacle avoidance path, and the total length of the obstacle avoidance path, and stored in the graph database Neo4j; Generate a case feature set based on the features of successful obstacle avoidance cases performed by the drone; Encrypt the generated case feature set; Send the encrypted case feature set to all nests and drones via wireless communication network; After receiving the case feature set, the machine nest and the drone perform decryption processing; The machine nest and drone are remotely updated based on the decrypted case feature set, and the updated results are fed back to the system to complete end-to-end technical iteration.

9. A drone inspection path optimization system based on a mobile drone nest, used to implement the method according to any one of claims 1 to 8, characterized in that: include: The global environmental situation map generation module is used to perform dynamic environmental perception and multi-source data fusion on the drone patrol area based on the mobile drone nest to generate a global environmental situation map; The task allocation matrix acquisition module is used to obtain inspection work orders and equipment defect history databases. Based on the global environment situation map, it performs intelligent task analysis and dynamic priority sorting to obtain a task allocation matrix with time windows and priorities. The dynamic flight path set output module is used to collect the performance parameters of the mobile drone nest, perform multi-objective hierarchical path planning based on the global environment situation map and task allocation matrix, and output a dynamic flight path set with a time stamp; The emergency self-healing system construction module is used to build a real-time obstacle avoidance and emergency self-healing system based on the dynamic flight path set, obtain a safe correction path set, and conduct drone inspections based on the safe correction path set; The global environment situation map update trigger module is used to obtain the remaining power of the machine nest and road network data, combine it with the safety correction path set, perform mobile machine nest coordinated scheduling and energy optimization, output the machine nest movement instruction set, and trigger the global environment situation map update; The knowledge evolution module is used to obtain drone execution logs, perform knowledge evolution, extract features of successful obstacle avoidance cases performed by drones, and complete end-to-end technology iteration.

Images