Multi-unmanned aerial vehicle oil depot inspection method based on hierarchical strategy

Abstract

The invention relates to the technical field of path planning, in particular to a multi-unmanned aerial vehicle oil depot storage tank inspection intelligent optimization method based on a layering strategy. The method comprises the following steps: constructing an equivalent terrain mathematical model according to oil depot storage tank characteristics, and establishing a multi-objective optimization objective function; performing global path planning inspection of multiple unmanned aerial vehicles by using an improved wolf pack optimization algorithm; abnormities are monitored through a carried sensor, and abnormal information is uploaded to a terminal; the terminal issues the target abnormal point to the single unmanned aerial vehicle; performing single-machine local path optimization by adopting an improved particle swarm optimization algorithm; and updating the state of the inspection task. According to the method, global path planning and local path optimization are effectively combined through a layered inspection strategy, the defects of a traditional method in inspection full coverage and local emergency response are overcome, the intelligent and efficient level of an unmanned aerial vehicle inspection system is improved, and the method is suitable for complex oil depot storage tank inspection tasks and emergency disposal scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) path planning, and in particular to a multi-UAV oil depot inspection method based on a hierarchical strategy. Background Art

[0002] The inspection task of oil depot tanks usually includes two aspects: comprehensive inspection and emergency response to emergencies. In the comprehensive inspection task, it is necessary to plan the global path of the drone cluster to cover all tank areas and conduct inspections as much as possible, while avoiding obstacles around the tanks to ensure the safety and efficiency of drone flight; in the event of emergencies such as fire, leakage or illegal intrusion, a single drone needs to quickly go to the target point for local inspection to check the abnormal points in detail and collect relevant data. This requires the path planning algorithm to ensure that the drone cluster can achieve efficient and comprehensive regional coverage in a complex environment, and to ensure that a single drone can respond quickly and arrive at the designated location accurately in an emergency. Traditional path planning algorithms such as A*, RRT and genetic algorithms (GA) have limitations in oil depot tank inspections. The A algorithm has strong path certainty but low efficiency and is difficult to adapt to complex environments; RRT* has limited path optimization capabilities and insufficient smoothness; GA converges slowly and is prone to local optimality. These defects make it difficult for traditional algorithms to simultaneously meet the dual needs of comprehensive inspection and emergency response to emergencies, posing a significant challenge to the efficiency and safety of inspection tasks.

[0003] In recent years, swarm intelligence optimization algorithms such as particle swarm optimization (PSO), ant colony optimization (ACO) and wolf pack optimization (GWO) have been widely used in path planning problems and have shown better results than traditional methods. Among them, the GWO algorithm has strong search capabilities and good real-time performance, which is suitable for the global path planning task of drone clusters. However, it is easy to fall into local optimality during local optimization, resulting in insufficient diversity of path node distribution; while the PSO algorithm, with its fast convergence characteristics and flexibility, can meet the needs of a single drone to quickly avoid obstacles and accurately reach the target point in a complex environment. However, it is difficult to simultaneously meet the dual requirements of global coverage and local efficient inspection in oil depot inspection when using these two algorithms alone. Therefore, according to the characteristics of oil depot tank inspection tasks, a multi-UAV oil depot inspection method based on a hierarchical strategy is proposed. Through the hierarchical optimization strategy of global and local path planning, the global coverage of the drone cluster and the precise inspection of a single drone are achieved, thereby taking into account the comprehensive coverage of the global inspection and the high efficiency of emergency handling, providing a safe, efficient and intelligent solution for oil depot tank inspection. Summary of the invention

[0004] The present invention aims to solve the efficiency and accuracy problems of multi-UAV global path planning and single UAV local path planning in oil depot tank inspection tasks, and provides a multi-UAV oil depot inspection method based on a hierarchical strategy. This method is suitable for comprehensive inspection and emergency handling scenarios of oil depot tanks, and can optimize resource scheduling and enhance system reliability while improving inspection efficiency and ensuring mission safety.

[0005] The multi-UAV oil depot inspection method based on hierarchical strategy is an intelligent planning method that focuses on both global coverage and local emergency response. It mainly includes two core task modules: global path planning and local path planning.

[0006] The method described has the following steps:

[0007] Step 1: Model construction:

[0008] (1) According to the distribution of oil tanks and surrounding terrain obstacles, the threat range and height limit of each tank are defined and abstracted into a threat equivalent terrain mathematical model to form the terrain threat area and the altitude limit of UAV flight.

[0009] (2) In combination with the requirements of UAV inspection tasks, an objective function for multi-objective optimization is designed to balance key indicators such as path length, threat avoidance, flight altitude, time synchronization, and collision cost.

[0010] Step 2: Global path planning:

[0011] (1) Using the improved wolf pack optimization algorithm to perform global path planning for multiple UAVs, the wolf pack positions are adjusted by simulating the hunting behavior of gray wolves to form an efficient global inspection path;

[0012] (2) Algorithm improvements include the introduction of dynamic inertia weights to enhance global search capabilities and convergence speed, and through fitness function optimization, comprehensive consideration of multiple constraints such as path length, static obstacle distribution, inspection coverage, and task time;

[0013] (3) After planning is completed, a global path suitable for multi-UAV collaborative operations is generated and assigned to each UAV to ensure a comprehensive inspection of the tank area.

[0014] Step 3: Real-time data upload and anomaly detection:

[0015] (1) When the drone performs the global path inspection task, it collects tank status information in real time through the onboard sensors and uploads it to the terminal system;

[0016] (2) The system detects abnormal conditions (such as fire, equipment failure, or illegal intrusion) in conjunction with inspection data;

[0017] (3) Wireless communication uploads abnormal information to the system terminal.

[0018] Step 4: Emergency information processing:

[0019] (1) Event data reception and analysis: The system terminal receives emergency event information uploaded by the drone, including the abnormal location, event nature, and related sensor data;

[0020] (2) Target point determination: The system analyzes the specific location and nature of the emergency, sets the coordinates of the target points that need to be checked, ensures that the target points are outside the obstacle area, and optimizes the task sequence according to the event priority;

[0021] (3) Task assignment: Based on the current inspection task status and UAV distribution, a suitable UAV is selected to perform the local path planning task, and the target point information is transmitted to the UAV.

[0022] Step 5: Local path planning:

[0023] (1) Initial particle swarm generation: Based on the received target point information, a group of particles (path schemes) are randomly initialized near the target area, and the fitness value is calculated for each particle, taking into account the path length, obstacle avoidance capability, and mission time;

[0024] (2) Particle speed and position update: By introducing a dynamic neighborhood selection mechanism, the speed and position of the particle are updated to avoid falling into the local optimal solution. At the same time, the particle constraints are corrected to ensure that its path plan does not cross obstacles and optimize the path smoothness.

[0025] (3) Path convergence and output: Continuously iterate and update the particle position until the convergence condition or the preset number of iterations is reached, and the optimal path is output as the local inspection route of a single UAV, and the UAV is assigned to quickly reach the target point;

[0026] Step 6: Update the inspection task status:

[0027] (4) Mission execution and feedback: The UAV reaches the target point according to the planned path, completes detailed inspection and mission response to emergencies, and feeds back inspection results and status information to the system terminal in real time, updates inspection records, and prepares to execute subsequent tasks.

[0028] The main features of the present invention include:

[0029] (1) The global and local planning have clear division of labor: the improved wolf pack algorithm is used for global path planning to adapt to different tank layouts and complex obstacle environments; the improved particle swarm algorithm is used for local path planning to quickly respond to emergencies. The two complement each other to form a complete path planning framework.

[0030] (2) Innovation of algorithm optimization:

[0031] The wolf pack algorithm significantly improves the efficiency and accuracy of global path planning through dynamic weights and multi-constraint fitness functions.

[0032] The particle swarm algorithm enhances the response speed and reliability of local path planning through dynamic neighborhood mechanism and update rule optimization.

[0033] (3) Multiple drones collaborate to complete global inspections, reducing the burden on a single drone; a single drone quickly responds to emergencies, avoiding mission delays and improving the safety and flexibility of inspection tasks.

[0034] (4) Optimize algorithm parameters for complex obstacle scenarios of tank distribution to ensure that the planning method is adaptable to tank layouts of different densities and complexities.

[0035] (5) Dynamic coordination between global and local planning is achieved through information interaction to ensure the consistency and efficiency of inspection tasks.

[0036] (6) Use intelligent algorithms to reduce path planning redundancy, shorten task time, and reduce energy consumption and equipment maintenance costs.

[0037] The technical solution of the present invention effectively improves the global coverage efficiency and local emergency response capability of UAV inspection tasks, and provides innovative ideas and practical basis for the intelligent application of UAVs in oil depots, storage tanks and other complex obstacle scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0039] Figure 1 It is a flow chart of the multi-UAV oil depot inspection method based on the hierarchical strategy.

[0040] Figure 2 It is a three-dimensional simulation diagram of the threat equivalent terrain of the present invention.

[0041] Figure 3 It is a two-dimensional schematic diagram of global path planning of the improved wolf pack optimization algorithm.

[0042] Figure 4 It is a two-dimensional schematic diagram of global path planning of the improved particle swarm optimization algorithm.

[0043] Figure 5 It is a three-dimensional schematic diagram of local path planning of the improved particle swarm optimization algorithm. DETAILED DESCRIPTION

[0044] The multi-UAV oil depot inspection method based on a hierarchical strategy and its implementation provided by the present invention are described in detail below in conjunction with the accompanying drawings.

[0045] The present invention proposes a path planning method based on bionic algorithm. Aiming at the inspection task of oil depot tanks, the path planning is divided into two main processes: one is to use the improved wolf pack optimization (GWO) algorithm to realize the global path planning of multi-UAV collaboration to ensure the coverage of the target area and optimize the inspection efficiency; the other is that when an emergency occurs, a single UAV performs detailed local path planning through the particle swarm optimization (PSO) algorithm to accurately inspect and respond to abnormal areas.

[0046] The overall control method process is as follows Figure 1 As shown below, combined Figure 1 The multi-UAV oil depot inspection method based on hierarchical strategy involved in the method is described in detail.

[0047] 1. Model construction:

[0048] Determine mission objectives, including tank inspection coverage, global path planning requirements, and local abnormal event response objectives.

[0049] like Figure 2 As shown, a three-dimensional model of the oil depot tank area is established, and the map data comes from the pre-processed oil depot terrain file.

[0050] Configure the multi-UAV collaborative system and define the initial position and starting task allocation plan for each UAV.

[0051] Equip the drone with high-precision sensors (such as LiDAR, cameras, IMU) and connect it to the terminal system for real-time data uploading and processing.

[0052] Initialize the communication network to ensure stable data interaction between the drone and the terminal system.

[0053] 2. Global path planning: improving the basic process of wolf pack optimization algorithm

[0054] Gray wolf behavior model: simulates gray wolf hunting behavior, maps the drone global path planning task into an optimization problem, and defines the search space and wolf individuals.

[0055] Initialize the wolf pack: randomly generate multiple wolf pack positions (i.e., global path planning schemes) and calculate the fitness value of each path.

[0056] Individual sorting: Sort by fitness value and determine three high-quality individuals α, β, and δ (representing the current optimal, second-optimal, and third-optimal paths, respectively).

[0057] Position update rule: Update the positions of the remaining wolves according to the positions of α, β, and δ:

[0058]

[0059] Among them, A and C are dynamic control parameters.

[0060] Iteration and convergence: Repeatedly update the position and calculate the fitness value until the convergence condition is met (such as reaching the preset number of iterations or the fitness no longer increases significantly).

[0061] Output the optimal path and assign it to the UAV, such as Figure 3 As shown, a global path planning solution for multi-UAV collaboration based on the improved wolf pack optimization algorithm is generated.

[0062] 3. Global Path Planning: Improvements in the Wolf Pack Optimization Algorithm

[0063] Introducing dynamic inertia weights to enhance the algorithm's global search capability and convergence speed:

[0064]

[0065] Where t is the current iteration number and T is the total iteration number. The dynamic weight is gradually reduced through iterations, the global search capability is enhanced in the early stage, and the convergence speed is optimized in the later stage.

[0066] Improved fitness function, taking into account path length, threat avoidance, mission time and collision cost:

[0067] f global =w1·f0+w2·f h +w3·f t +w4·f m +w5·f c

[0068] f0 is the fuel consumption cost (normalized to represent the total range of each UAV).

[0069] f h Altitude limit penalty (accumulated penalties for exceeding or falling below the altitude limit).

[0070] f t Threat avoidance cost (threat index considering radar threat and other threats).

[0071] f m Task time synchronization cost (the absolute value of the difference between task time and global time synchronization).

[0072] f c Collision cost (total number of collisions between drones).

[0073] Among them, the weight w = [w1, w2, w3, w4, w5] controls the relative importance of each sub-goal.

[0074] Added multi-constraint processing mechanism to ensure that the path avoids obstacles and achieves balanced task distribution:

[0075] The path is evaluated through the trajectory detection function to determine whether there are any of the following problems:

[0076] Trajectory problem points: Are there points where the path does not meet the constraints?

[0077] Threat area: whether entering radar or other threat areas;

[0078] Task time allocation: Ensure that each drone completes its task in a balanced time to avoid delays.

[0079] 4. Real-time data upload and anomaly detection

[0080] During the global path inspection, drones collect tank status information in real time, including equipment appearance, thermal imaging data, and gas detection results.

[0081] The data is uploaded to the terminal system through the wireless communication module.

[0082] The terminal performs anomaly detection on the uploaded data to identify situations such as fire, equipment failure or illegal intrusion.

[0083] Abnormal event information, including location, nature and related sensor data, is recorded to the system terminal.

[0084] 5. Emergency information processing

[0085] Event reception and analysis: The terminal system receives abnormal information uploaded by the drone and analyzes the specific location and nature of the emergency.

[0086] Target point determination: Determine the target points that need to be checked based on the analysis results, ensure that obstacles are avoided, and optimize the processing order according to task priority.

[0087] Task assignment: Consider the current UAV position and task status and select the most suitable UAV to perform the local path planning task.

[0088] Mission upload: transmit the target point information and mission instructions to the designated drone. Figure 4 This is a two-dimensional schematic diagram of a single UAV path generated based on the improved particle swarm optimization algorithm. Figure 5 is the corresponding three-dimensional schematic diagram.

[0089] 6. Local path planning: basic process of particle swarm optimization algorithm

[0090] Particle initialization: Taking the target point as a reference, a particle swarm is randomly generated in the target area, and each particle represents a local path plan.

[0091] Fitness function calculation: The fitness function calculation method consistent with the wolf pack optimization algorithm in global path planning is adopted, which comprehensively considers path length, threat avoidance, task time and collision cost.

[0092] Speed ​​and position update: Update particle speed and position according to the following formula:

[0093] v i (t+1)=w·v i (t)+c1·r1·(p best -x i )+c2·r2·(g best -x i )

[0094] x i (t+1)=x i (t)+v i (t+1)

[0095] v i (t): velocity of particle i at the tth iteration.

[0096] x i (t): The position of particle i at the tth iteration.

[0097] p best : The historical optimal position of particle i.

[0098] g best : The global optimal position of the particle swarm.

[0099] w, c1, c2, r1, r2: inertia weights and acceleration factors.

[0100] Individual optimality and global optimality: Update the historical optimal position p of each particle best and the global optimal position of the particle swarm g best .

[0101] Iteration and Convergence: Repeatedly update particle velocity and position until convergence conditions or the preset maximum number of iterations are reached.

[0102] Output the optimal path: The path corresponding to the global optimal position of the particle swarm is used as the local inspection route.

[0103] 7. Local path planning: improvements of particle swarm optimization algorithm

[0104] During the operation of the algorithm, the learning factors c1 and c2 are dynamically adjusted. In the initial stage, c1 is made higher and c2 is made lower to enhance the exploration ability. In the later stage, c1 is gradually reduced and c2 is increased to accelerate the convergence to the global optimal solution.

[0105]

[0106] Among them, t is the current iteration number and T is the maximum iteration number.

[0107] Introduce the rebound method for boundary processing to improve search efficiency:

[0108] When the particle position exceeds the boundary, it is adjusted to a position symmetrical to the boundary in the search space through the rebound method to retain the diversity of particles near the boundary and avoid information loss.

[0109] For particles beyond the maximum bounds:

[0110] x i '=x max -(x i -x max )

[0111] For particles beyond the minimum boundary:

[0112] x i '=x min +(x min -x i )

[0113] Introduce early stopping mechanism to improve computing efficiency:

[0114] By monitoring the change of the optimal fitness value of the particle swarm, when the change is less than the preset threshold ε (i.e. g best (t+1)-g best (t)|<ε), the iteration is terminated early to reduce invalid calculations and improve real-time planning performance.

[0115] Adopting grid-based statistical optimization method to improve global search performance:

[0116] The search space is divided into grid areas, the number of particles in each grid is counted, and the particle-dense areas are searched first to avoid redundant calculations. This method is combined with the early stopping mechanism to effectively reduce the computational burden.

[0117] 8. Abnormal point inspection process

[0118] Data collection: After the drone reaches the abnormal point, it uses sensors to carefully scan the equipment status and collect high-resolution images, gas concentration data and temperature data.

[0119] Data feedback and analysis: The collected inspection data is uploaded to the terminal system through the communication module for detailed threat assessment.

[0120] 9. Update inspection task status

[0121] After completing the local inspection task, update the task status:

[0122] Task completion mark: Mark the completed inspection points and update the inspection coverage data in the global path planning.

[0123] Return to the initial position or relay point: After completing a local mission, the drone returns to the original mission area or goes to a relay point to replenish power and data.

[0124] Path correction and replanning: Based on the current status of the UAV that has completed the task (battery power, load, etc.), it is assigned a new inspection task and the GWO algorithm is re-executed to update the global path if necessary.

[0125] Queue adjustment: If a drone is unable to continue to perform a mission (e.g., insufficient battery), the task queue is dynamically adjusted and assigned to a backup drone.

[0126] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A multi-UAV oil depot inspection method based on a hierarchical strategy, characterized in that: include: S1. Model construction: According to the distribution of oil tanks and surrounding terrain obstacles, a threat equivalent terrain mathematical model is constructed to define the threat range and flight altitude limit of the tanks. In combination with the inspection task requirements, a multi-objective optimization objective function is designed to comprehensively balance key indicators such as path length, threat avoidance, flight altitude, time synchronization and collision cost. S2. Global path planning: an improved wolf pack optimization algorithm is used to generate a global path planning scheme for the drone cluster. By initializing the population, updating the position and calculating the fitness, a reference path for cluster inspection is generated in combination with path smoothing and boundary processing mechanisms. S3. Data upload and anomaly detection. During the global path inspection mission, the drone collects tank status data in real time through onboard sensors and uploads it to the terminal system. The system detects abnormal situations (such as fire, equipment failure or illegal intrusion) based on the inspection data and transmits the abnormal information to the terminal through wireless communication. S4, emergency information processing, the system terminal receives emergency information uploaded by the drone (including abnormal location, event nature and related sensor data), analyzes the specific location and nature of the event, determines the coordinates of the target point outside the obstacle area, and optimizes the task sequence according to the event priority. Combined with the current inspection task status and drone distribution, the system selects a suitable drone to perform the local path planning task and transmits the target point information to the drone. S5. Local path planning. For the location of abnormal points, a single UAV uses an improved particle swarm optimization algorithm to calculate the local optimal path, avoid obstacles, accurately inspect abnormal points and record relevant information. By initializing the particle position and parameters, combining boundary processing and dynamic learning factor adjustment, it ensures that the path generation meets the fitness requirements, and finally outputs the local path optimal solution to guide the UAV to complete the task. S6. Update the inspection task status. After completing all inspection tasks, upload the inspection results and abnormal point processing data to the terminal, and update the task status to ensure data integrity.

2. The multi-UAV oil depot inspection method based on a hierarchical strategy according to claim 1 is characterized in that: The steps of improving the wolf pack optimization algorithm include: (1) Initialize the wolf pack: randomly generate multiple initial paths (i.e., wolf pack locations) and define a fitness function to evaluate path performance; (2) Path smoothness optimization: The initial path is smoothed by interpolation to improve the continuity and feasibility of the path; (3) Boundary processing mechanism: The rebound method is used to adjust the path points beyond the boundary to ensure the integrity of the search space; (4) Individual sorting and leader selection: Sort the wolf pack individuals according to their fitness values ​​and select three leader individuals α, β, and δ (optimal, suboptimal, and third-optimal paths); (5) Dynamic parameter adjustment: Dynamically adjust control parameters by combining the number of iterations and nonlinear decreasing strategy to balance global exploration and local development; (6) Position update strategy: Based on the positions of α, β, and δ, the positions of other individuals are updated in combination with the obstacle cost gradient to ensure the rationality and safety of the path; (7) Iteration and path output: Repeatedly update the wolf pack position and fitness value until the convergence condition or the maximum number of iterations is reached, and output the global optimal path to allocate to the UAVs, forming a global path planning solution for multi-UAV collaboration.

3. The multi-UAV oil depot inspection method based on a hierarchical strategy according to claim 1 is characterized in that: The improved particle swarm optimization algorithm comprises the following steps: (1) Particle initialization: Taking the target point as a reference, randomly generate a particle swarm in the target area, initialize the particle speed and position, and define the fitness function; (2) Speed ​​and position update: The speed and position of the particle are calculated based on the particle’s current speed, individual historical optimal position, and global optimal position, combined with the inertia weight and acceleration factor, and the rebound method is used to adjust the particles that exceed the boundary; (3) Dynamic learning factor adjustment: Dynamically adjust the learning factor according to the number of iterations to balance global exploration and local development and improve search efficiency; (4) Individual optimal and global optimal update: Real-time update of the historical optimal position of each particle and the global optimal position of the particle group; (5) Early stopping mechanism and path output: When the change in the optimal fitness value is less than the set threshold or the maximum number of iterations is reached, the iteration is stopped and the path corresponding to the global optimal position is used as the local inspection route.

4. The multi-UAV oil depot inspection method based on a hierarchical strategy according to claim 1 is characterized in that: The multi-objective optimization objective function consists of the following cost terms: (1) f0 fuel consumption cost (normalized to represent the total flight range of each UAV). (2)f h Altitude limit penalty (accumulated penalties for exceeding or falling below the altitude limit). (3)f t Threat avoidance cost (threat index considering radar threat and other threats). (4)f m Task time synchronization cost (the absolute value of the difference between task time and global time synchronization). (5)f c Collision cost (total number of collisions between drones). The objective function is formed by the weighted sum of each cost term, and the formula is as follows: <h2 style=";text-align:left;direction:ltr">f<h2 style=";text-align:left;direction:ltr"> global <h2 style=";text-align:left;direction:ltr"> = w1 f0 + w2 f<h2 style=";text-align:left;direction:ltr"> h <h2 style=";text-align:left;direction:ltr"> +w3·f<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> +w4·f<h2 style=";text-align:left;direction:ltr"> m <h2 style=";text-align:left;direction:ltr"> +w5·f<h2 style=";text-align:left;direction:ltr"> c <h2 style=";text-align:left;direction:ltr"> (1) Among them, the weight w = [w1, w2, w3, w4, w5] controls the relative importance of each sub-goal.

Images