Inspection route planning method and system for hydrogen-powered drones

Abstract

This application relates to the field of image processing technology, specifically to a method and system for planning inspection routes for hydrogen-powered drones. The method includes: calculating RSSI values ​​at various locations within the inspection area using a signal attenuation model, as simulated RSSI values; using a tensor constructed from these simulated values ​​as a benchmark, correcting the tensor based on actual measurements from each inspection; constructing a first optimized feature value based on the similarity of the corrected tensors from historical inspections and the distribution characteristics of RSSI values ​​within the inspection area; determining the size of the optimization area using the first optimized feature value and the inspection route length; and optimizing the corresponding routes within the optimization area using an optimization algorithm to obtain the optimal route. This improves the inspection performance of hydrogen-powered drones, enabling them to better adapt to changes in cellular communication signal strength during inspections, ensuring data transmission stability and efficiency, and enhancing overall inspection efficiency and quality.

Description

Technical Field

[0001] This application relates to the field of image processing technology, specifically to a method and system for inspection route planning for hydrogen-powered drones. Background Technology

[0002] With the development of hydrogen-powered drone technology, it has been widely used in various inspection fields, especially for the inspection of outdoor power transmission networks. Compared with other types of drones, hydrogen-powered drones have significant advantages such as longer endurance and higher payload, and can carry high-resolution cameras and other monitoring equipment to collect richer and more detailed data. However, this also leads to higher requirements for data transmission performance.

[0003] However, current path planning methods mainly focus on factors such as path length and altitude cost, while ignoring the data transmission performance of drone inspections. Since the distance between different locations in the inspection area and the base station varies, the signal strength received at different locations also varies. The data transmission quality of drones in locations with poor signal will be greatly affected, thus affecting the inspection effect of hydrogen-powered drones. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a method and system for inspection route planning of hydrogen-powered drones, and the specific technical solution adopted is as follows:

[0005] In a first aspect, embodiments of this application provide a method for planning inspection routes for hydrogen-powered unmanned aerial vehicles (UAVs), the method comprising the following steps:

[0006] Obtain the inspection route and take-off and landing points of the UAV during its first inspection within the inspection area; construct a three-dimensional grid area between two adjacent take-off and landing points; combine the signal attenuation model and the signal transmission frequency of the base station in the inspection area to calculate the RSSI simulation value of each grid in each grid area and construct the three-dimensional RSSI simulation tensor of each grid area.

[0007] The grids on the inspection route are recorded as route grids; based on the measured RSSI values ​​of all route grids in each grid area during all historical inspections, the RSSI simulation tensor is corrected to obtain the RSSI correction tensor of each grid area after the current inspection and the RSSI correction value of each grid.

[0008] For each grid region, based on the similarity between the RSSI correction tensors of all adjacent grid regions after two historical inspections, the stable characteristic value of the RSSI distribution of the current grid region after inspection is calculated.

[0009] Based on the differences in RSSI correction values ​​and positions between each route grid and its corresponding neighboring grid within each grid area during the initial inspection, and combined with the RSSI distribution stability feature value, the first optimized feature value for each grid area in each inspection is constructed.

[0010] For each grid area inspected, based on the number of route grids within the grid area and the first optimization feature value, the inspection route within the grid area is segmented and the optimization area corresponding to each segment of the inspection route is determined; based on the RSSI correction value distribution characteristics of the grids in each optimization area, combined with the optimization algorithm, the route of each inspection route segment is optimized.

[0011] In one embodiment, the process of obtaining the three-dimensional RSSI simulation tensor of each grid region is as follows:

[0012] Two adjacent take-off and landing points are respectively used as the two diagonal vertices of a cuboid grid region to construct a three-dimensional grid region between the two adjacent take-off and landing points.

[0013] The location and signal transmission power of all base stations within the inspection area are used as inputs to the signal attenuation model. The RSSI value of each grid in each grid area is calculated and recorded as the simulated RSSI value of each grid.

[0014] The third-order tensor composed of the RSSI simulation values ​​of all grid cells within each grid region is denoted as the three-dimensional RSSI simulation tensor of each grid region.

[0015] In one embodiment, the process of obtaining the RSSI correction tensor of each grid region after the current inspection and the RSSI correction value of each grid is as follows:

[0016] The simulated RSSI tensor of each grid area and the measured RSSI values ​​of all route grids in each grid area during the first inspection are used as inputs to the Bayesian calibration method. The output is the corrected RSSI tensor of each grid area after the first inspection, which is denoted as the RSSI correction tensor.

[0017] For each grid area, the measured RSSI values ​​of all route grids within the current inspection area and the RSSI correction tensor of the grid area after the previous inspection are used as inputs to the Bayesian calibration method, and the output is the RSSI correction tensor of the grid area after the current inspection.

[0018] Each element in the RSSI correction tensor of the current inspected grid area is recorded as the RSSI correction value of each grid cell in the current inspected grid area.

[0019] In one embodiment, the process of obtaining the stable characteristic value of the RSSI distribution is as follows:

[0020] For each grid region, the similarity between the RSSI correction tensor of each inspection and the grid region after the previous inspection is denoted as the first similarity of the inspected grid region; the stable eigenvalue of the RSSI distribution of the current inspected grid region is denoted as... , The expression is:

[0021]

[0022] in, This indicates the number of pre-selected inspections prior to the current inspection. This represents the first similarity of the current grid region in the nth inspection prior to the current inspection; This is the normalization function.

[0023] In one embodiment, the process of obtaining the first optimized feature value is as follows:

[0024] During the initial inspection, in each grid area, the grid with an RSSI correction value greater than that of any given grid is considered a strong signal grid in the neighborhood of any given grid.

[0025] Calculate the first optimized feature value of the current grid region being inspected. , The expression is:

[0026]

[0027] in, This represents the stable RSSI distribution characteristic value of the current grid area being inspected. This indicates the number of route grid cells within the current grid area during the initial inspection; R represents the number of strong signal grids in the neighborhood of the i-th route grid within the current grid area during the first inspection; R represents the neighborhood radius of the preset route grid. This represents the distance between the i-th route grid cell within the current grid area during the first inspection and the j-th strong signal grid cell in its neighborhood; This represents the RSSI correction value of the i-th route grid within the current grid area during the first inspection; This represents the difference between the RSSI correction value of the i-th route grid within the current grid area during the first inspection and the j-th strong signal grid in its neighborhood; This is the normalization function.

[0028] In one embodiment, the step of segmenting the inspection route within the grid area and determining the optimization area corresponding to each segment of the inspection route specifically involves:

[0029] Within the current grid area of ​​the current inspection, calculate the product of the number of route grids and the corresponding first optimization feature value, and record it as the first product; construct the first optimization area with the starting point of the inspection route in the grid area as the center and the rounded-up result of the first product as the radius; take the route grid farthest from the starting point in the inspection route after the starting point in the first optimization area as the endpoint in the first optimization area, and obtain the inspection route between the starting point and the endpoint in the first optimization area;

[0030] The endpoint of the first optimized region is used as the starting point of the next optimized region. The same acquisition method as the first optimized region and its endpoint is used to construct the next optimized region and its endpoint, thus obtaining the inspection route between the starting point and the endpoint in the next optimized region. This process is repeated until the endpoint of the optimized region becomes the endpoint of the inspection route in the current grid region being inspected, at which point the construction of a new optimized region stops. Each inspection route is the inspection route between the starting point and the endpoint in each optimized region.

[0031] In one embodiment, the route optimization for each inspection route segment specifically includes:

[0032] For any optimization region, calculate the mean of the RSSI correction values ​​of all route grids in the optimization region, and record it as the first mean; mark the grids in the optimization region that are less than the first mean as candidate obstacle grids, and sort all candidate obstacle grids in the optimization region according to their three-dimensional coordinates, and obtain the sequence number of each candidate obstacle grid in the optimization region based on the sorting result;

[0033] The vector composed of the indices of a predetermined number of candidate obstacle grids randomly selected from the optimization region in any given time is denoted as the obstacle vector selected in that time. All obstacle grids corresponding to the obstacle vector, as well as the start and end points in the optimization region, are taken as A. The algorithm's input and the output optimized route serve as the optimized route corresponding to the obstacle vector;

[0034] The optimized route corresponding to any selected obstacle vector is recorded as the current optimized route. The difference between the current optimized route and the corresponding initial inspection route is analyzed, and the second optimized feature value of the current optimized route is constructed.

[0035] For each particle in the particle swarm optimization algorithm, the obstacle vector randomly obtained each time is taken as a particle; the second optimization feature value of the optimization route corresponding to each particle is taken as the fitness of each particle. The particle swarm optimization algorithm is used to optimize the inspection route between the starting point and the ending point in the optimization region and output the optimal route.

[0036] In one embodiment, the process of obtaining the second optimized feature value is as follows:

[0037] Obtain the initial route corresponding to each current optimized route; calculate the second optimization feature value S of the current optimized route, where S is expressed as:

[0038]

[0039] in, This represents the absolute value of the difference between the number of grid cells on the current optimized route and the corresponding initial route; Indicates the number of grid cells in the optimized route; This represents the RSSI correction value at the m-th grid on the current optimized route; This represents the RSSI correction value of the grid cell that is closest to the m-th grid cell on the current optimized route, on the initial route corresponding to the current optimized route. This represents the distance between the m-th grid on the current optimized route and its nearest corresponding grid on the initial route; The minimum positive number is preset. This represents an exponential function with the natural constant as its base.

[0040] In one embodiment, the process of obtaining the initial route corresponding to each current optimized route is as follows:

[0041] The points closest to the start and end points of the current optimized route in the initial inspection route are taken as the corresponding points of the start and end points, respectively. The route between the corresponding points of the start and end points in the initial inspection route is taken as the initial route corresponding to the current optimized route.

[0042] Secondly, embodiments of this application also provide an inspection route planning system for hydrogen-powered drones, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0043] The embodiments of this application have at least the following beneficial effects:

[0044] This application calculates the RSSI values ​​at various locations within the inspection area using a signal attenuation model, serving as simulated RSSI values. A tensor constructed from these simulated values ​​is used as a benchmark, and calibration is performed using actual RSSI measurements at each location along the inspection route during each inspection. This dynamic updating during each inspection more accurately reflects the wireless communication signal strength distribution within the inspection area. The calculation of stable RSSI distribution characteristics improves the accuracy of distribution network cellular network assessment, providing more reliable data support for inspection route planning. During the planning of hydrogen-powered drone inspection routes, a first optimization characteristic value is constructed, comprehensively considering signal strength and the validity of inspection data. When optimizing the route in segments, the size of the optimization area can be adaptively adjusted, allowing for more precise adaptation to changes in signal strength within the area. The size of the optimization area is determined by the first optimization characteristic value and the inspection route length. For the corresponding routes requiring optimization within the optimization area, A... The inspection route is optimized using algorithms and particle swarm optimization to obtain the optimal route. Inspection is then carried out according to the optimal route, which effectively improves the inspection performance of the hydrogen-powered drone. This allows it to better adapt to changes in cellular communication signal strength during the inspection process, ensuring the stability and efficiency of data transmission, while reducing energy consumption, extending flight time, and improving overall inspection efficiency and quality. Attached Figure Description

[0045] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating the steps of an inspection route planning method for hydrogen-powered drones according to an embodiment of this application;

[0047] Figure 2 This is a schematic diagram illustrating the process of obtaining RSSI correction values. Detailed Implementation

[0048] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive objective, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the inspection route planning method and system for hydrogen-powered drones proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0050] The following, in conjunction with the accompanying drawings, provides a detailed explanation of the specific scheme for the inspection route planning method and system for hydrogen-powered drones provided in this application.

[0051] Please see Figure 1 The diagram illustrates a flowchart of a method for planning inspection routes for hydrogen-powered drones according to an embodiment of this application. The method includes the following steps:

[0052] Step S1: Obtain the inspection route and take-off and landing points of the UAV during its first inspection within the inspection area; construct a three-dimensional grid area between two adjacent take-off and landing points; combine the signal attenuation model and the signal transmission frequency of the base station within the inspection area to calculate the RSSI simulation value of each grid in each grid area, and construct the three-dimensional RSSI simulation tensor of each grid area.

[0053] There are multiple take-off and landing points in the inspection route of the power distribution network inspection area. The setting of the take-off and landing points is affected by the specific requirements of the power distribution network inspection task, such as the scope of the inspection area, the distribution of transmission lines and the conditions of the take-off and landing site. Determining the take-off and landing points is to enable the drone to efficiently cover the entire inspection area and complete various inspection tasks.

[0054] This application obtains the initial inspection route and the locations of each take-off and landing point along the route based on the inspection tasks of the power distribution network inspection area and the flight parameters of the hydrogen-powered drone. The process of obtaining the initial inspection route is a well-known technique and will not be described in detail here. The initial inspection route is the route taken during the first inspection.

[0055] Each pair of adjacent take-off and landing points is designated as the two diagonally opposite vertices of a cuboid grid region to construct a three-dimensional grid region between them. This grid region completely encompasses the inspection route between the two adjacent take-off and landing points, with each grid cell being a cube with a side length of 1m. By constructing the three-dimensional grid regions for each initial inspection sub-route, a basic spatial structure is provided for subsequent inspection route planning. It should be noted that the size of a single grid cell can be set by the implementer according to the actual situation; this application does not impose specific limitations.

[0056] The location and signal transmission power of all base stations within the distribution network inspection area are obtained and used as input to a signal attenuation model. The model analyzes the signal propagation loss in three-dimensional space and calculates the RSSI (Received Signal Strength Indication) value for each grid cell in each grid area, recording it as the simulated RSSI value for each grid cell. The calculation of the RSSI value at each location using the signal attenuation model is a well-known technique, and the specific process will not be elaborated further. This embodiment uses the Okumura-Hata model as the signal attenuation model. It should be noted that many existing signal attenuation models exist, and implementers can use other signal attenuation models to calculate the RSSI values ​​at each location within the distribution network inspection area; this application does not impose any specific limitations.

[0057] For each three-dimensional grid region, the three-dimensional RSSI simulation tensor of the grid region is denoted as the third-order tensor formed by the RSSI simulation values ​​of all grids in the grid region according to the grid spatial position arrangement, which reflects the wireless communication signal strength distribution of the inspection area under the signal attenuation model.

[0058] Step S2: Record the grids on the inspection route as route grids; based on the measured RSSI values ​​of all route grids in each grid area during all historical inspections, correct the RSSI simulation tensor to obtain the RSSI correction tensor of each grid area after the current inspection and the RSSI correction value of each grid.

[0059] The RSSI simulation tensor is dynamically corrected using the actual RSSI values ​​measured during each inspection to improve its prediction accuracy in real-world environments. Specifically:

[0060] During each inspection, the RSSI value of each grid on the inspection route can be obtained by actual measurement. During each inspection, the grids in each grid area that are on the inspection route are recorded as route grids, and the actual measured RSSI value at each route grid is recorded as the RSSI measured value of that route grid.

[0061] For each grid area, the RSSI simulation tensor of the grid area and the measured RSSI values ​​of all route grids in the grid area during the first inspection are used as inputs to the Bayesian calibration method to correct the RSSI simulation tensor. The output is the corrected RSSI tensor of the grid area after the first inspection, which is denoted as the RSSI correction tensor.

[0062] Since the inspection route between two adjacent take-off and landing points may change through path optimization during each inspection, the inspection route within each grid area may differ between different inspections. The measured RSSI values ​​of all grids along the inspection route within each grid area during each inspection, along with the RSSI correction tensor of that grid area after the previous inspection, are used as inputs to the Bayesian calibration method. This method performs a new correction on the RSSI correction tensor of that grid area after the previous inspection, and outputs the RSSI correction tensor of that grid area after the current inspection. The Bayesian calibration method is a well-known technique, and its specific process will not be elaborated further.

[0063] This yields the RSSI correction tensor for each grid region after each inspection, and each element in the RSSI correction tensor is recorded as the RSSI correction value of each grid cell in each grid region after each inspection.

[0064] Step S3: For each grid region, based on the similarity between the RSSI correction tensors of all adjacent grid regions after two previous inspections, calculate the stable characteristic value of the RSSI distribution of the current grid region after inspection.

[0065] Taking any grid region as the current grid region, and using the current grid region after the current inspection as an example, the RSSI correction tensor of the grid region after the previous N inspections is obtained. If there are fewer than N inspections, Newton's interpolation method is used to supplement the missing RSSI values, thereby supplementing the RSSI correction tensor for N inspections. In this embodiment, the value of N is set to 10. In other embodiments of this application, the implementer can set the value of N according to the actual situation.

[0066] The similarity between the RSSI correction tensor of each inspection and the same grid region after the previous inspection is calculated and denoted as the first similarity of the grid region in each inspection. It should be noted that this application uses the cosine similarity between RSSI correction tensors as the similarity between RSSI correction tensors. There are many existing methods for calculating similarity, and implementers may also use other similarity algorithms to obtain the similarity between RSSI correction tensors. This application does not impose any specific restrictions.

[0067] The expression for calculating the stable characteristic value of RSSI distribution in a raster region is as follows:

[0068]

[0069] in, This represents the stable RSSI distribution characteristic value of the current grid area being inspected. This indicates the number of pre-selected inspections prior to the current inspection. This represents the first similarity of the current grid region in the nth inspection prior to the current inspection; The purpose of this normalization function is to limit the range of TV. In this embodiment, the arctangent normalization function is used to normalize the TV range. Normalize to the interval (0,1). In other embodiments of this application, the implementer may also use other normalization methods to... Normalize.

[0070] The greater the cosine similarity between the RSSI tensors before and after correction, the smaller the change in the distribution of cellular communication signal intensity in the grid area within two adjacent inspection intervals; This reflects the stability of the overall cellular communication signal strength in the grid area over time. The larger the distribution stability characteristic value, the smaller the change in the cellular communication signal strength distribution in the grid area over time, indicating that the current corrected signal strength data is more reliable, and the higher the reliability of subsequent route optimization based on the RSSI value.

[0071] Step S4: Based on the difference in RSSI correction value and position difference between each route grid and its corresponding neighboring grid in each grid area during the first inspection, and combined with the RSSI distribution stability feature value, construct the first optimized feature value for each grid area in each inspection.

[0072] During the initial inspection, in the current grid area, for each route grid on the inspection route, a neighborhood of each route grid is constructed with each route grid as the center and a preset length R as the radius. Grids within this neighborhood are then recorded as neighboring grids of that route grid. In this embodiment, the value of R is set to 5 grids. In other embodiments of this application, the implementer can set the value of R according to actual conditions.

[0073] Furthermore, taking the i-th route grid as an example, grids in its neighborhood with RSSI correction values ​​greater than the RSSI correction value of the i-th route grid are considered as strong signal grids in that neighborhood.

[0074] Based on the RSSI value distribution within the grid area of ​​the previous inspection and the difference in RSSI correction values ​​between the route grid and the corresponding strong signal grid in the neighborhood during the first inspection, the first optimized feature value of the grid area for each inspection is calculated, expressed as:

[0075]

[0076] in, The first optimized feature value of the current grid region being inspected; This represents the stable RSSI distribution characteristic value of the current grid area being inspected. This indicates the number of route grid cells within the current grid area during the initial inspection; R represents the number of strong signal grids in the neighborhood of the i-th route grid within the current grid area during the first inspection; R represents the neighborhood radius of the preset route grid. This represents the distance between the i-th route grid cell within the current grid area during the first inspection and the j-th strong signal grid cell in its neighborhood; This represents the RSSI correction value of the i-th route grid within the current grid area during the first inspection; This represents the difference between the RSSI correction value of the i-th route grid within the current grid area during the first inspection and the j-th strong signal grid in its neighborhood; The purpose of this normalization function is to control the calculation range of F. In this embodiment, the arctangent normalization function is used. Normalize to the interval (0,1).

[0077] A larger TV (Transmission Value) indicates a smaller variation in signal strength within the cellular communication area. This results in greater reliability of route optimization based on the corrected RSSI value, a smaller first optimization feature value, and a smaller subsequent optimization area. When optimizing routes in segments, optimizing routes within smaller optimization areas can more accurately adapt to changes in signal strength within the area, exhibiting greater adaptability and flexibility.

[0078] The smaller the value, the closer the grid with the strongest signal is to the initial route, and the greater the flexibility of subsequent route optimization. A smaller optimization area can be set to obtain a better inspection route.

[0079] The larger the value, the stronger the signal strength of the grid within the neighborhood. The more significant the improvement in the information transmission performance of the hydrogen-powered drone after route optimization, the more effective the optimization area becomes. Setting a smaller optimization area allows for full utilization of grids with stronger signal strength, resulting in a better inspection route. (Ratio) This is to obtain the relative improvement of the route raster RSSI value relative to the initial route; It reflects the average optimizability of the route grid from which the initial route was obtained.

[0080] F reflects the feasibility of route optimization within the grid area. A smaller first optimization eigenvalue corresponds to a stronger cellular signal strength in the area surrounding the initial route within the grid area, and a closer proximity to the signal strength. Setting a smaller optimization area allows for better adaptation to signal strength variations, resulting in a more efficient inspection route for the hydrogen-powered UAV. Simultaneously, a smaller first optimization eigenvalue corresponds to smaller variations in cellular communication signal strength within the grid area, leading to greater reliability in route optimization based on the corrected RSSI value. When optimizing a route in segments, optimizing within a smaller optimization area allows for more precise adaptation to changes in signal strength within the area, exhibiting greater adaptability and flexibility.

[0081] Step S5: For each grid area inspected, based on the number of route grids in the grid area and the first optimization feature value, the inspection route in the grid area is segmented and the optimization area corresponding to each segment of the inspection route is determined; based on the RSSI correction value distribution characteristics of the grids in each optimization area, combined with the optimization algorithm, the route of each inspection route segment is optimized.

[0082] Calculate the number of route grids in the current grid area of ​​the current inspection, denoted as L; calculate the product between L and the first optimized feature value F of the current grid area of ​​the current inspection, denoted as the first product.

[0083] For the inspection route of the current grid area being inspected, the first optimization area is constructed with the starting point of the inspection route as the center and the rounded-up result of the first product as the radius. The optimization area is spherical. In the inspection route after the starting point in the optimization area, the route grid farthest from the starting point is obtained as the endpoint in the optimization area. The inspection route between the starting point and the endpoint in the optimization area is optimized based on the data in the optimization area.

[0084] Using the endpoint as the next starting point, and still using the first product as the radius, construct the next optimization region. Similarly, within the next optimization region, take the farthest route grid on the inspection route after the next starting point as the endpoint of the next optimization region, and optimize the inspection route between the starting point and the endpoint in the next optimization region. Repeat this process until the endpoint of an optimization region becomes the endpoint of the inspection sub-route, at which point the construction of a new optimization region stops.

[0085] Specifically, the inspection route between the start and end points in each optimization region is taken as the inspection route for each segment; the inspection route for each segment is optimized based on the RSSI correction value distribution characteristics of the grid in each optimization region, as follows:

[0086] For any optimization region, (1) calculate the mean of the RSSI correction values ​​of all route grids in the optimization region, and denot it as the first mean; mark the grids in the optimization region that are less than the first mean as candidate obstacle grids, and sort all candidate obstacle grids in the optimization region according to the three-dimensional coordinates. When sorting, first compare the x coordinates, and the smaller x is placed first; if the x is the same, then compare the y coordinates, and the smaller y is placed first; if both x and y are the same, then compare the z coordinates, and the smaller z is placed first; thus, the sequence number of each candidate obstacle grid in the optimization region is obtained according to the sorting result.

[0087] (2) Randomly select a preset number of candidate obstacle grids in the optimization area, and record the vector composed of the serial numbers of the preset number of candidate obstacle grids selected in this application as the obstacle vector selected in this application. In this application, the preset number is set to half of the number of grids on the initial inspection sub-route in the optimization area; all obstacle grids corresponding to the obstacle vector selected in this application, as well as the start and end points in the optimization area, are taken as A. The algorithm takes as input an optimized route between the start and end points of the optimization region, and outputs the optimized route corresponding to the selected obstacle vector. Where A... The algorithm is a well-known technology, and the specific process will not be described in detail.

[0088] (3) For any selected obstacle vector, the optimized route is recorded as the current optimized route. The difference between the current optimized route and the corresponding initial inspection route is analyzed, and feature values ​​are constructed:

[0089] First, the points closest to the start and end points of the current optimized route in the initial inspection route are taken as the corresponding points of the start and end points, respectively. The route between the corresponding points of the start and end points in the initial inspection route is taken as the initial route corresponding to the current optimized route.

[0090] Then, the second optimization feature value of the current optimized route is calculated, expressed as:

[0091]

[0092] in, This represents the second optimization eigenvalue of the current optimization route; This represents the absolute value of the difference between the number of grid cells on the current optimized route and the corresponding initial route; Indicates the number of grid cells in the optimized route; This represents the RSSI correction value at the m-th grid on the current optimized route; This represents the RSSI correction value of the grid cell that is closest to the m-th grid cell on the current optimized route, on the initial route corresponding to the current optimized route. This represents the distance between the m-th grid on the current optimized route and its nearest corresponding grid on the initial route; The minimum positive number is preset to prevent the denominator from being zero; This represents an exponential function with the natural constant as its base, and its purpose is to reflect... and The numerical change relationship. Preferably, in the embodiments of this application, the numerical change relationship is... The value is set to 1. In other embodiments of this application, the implementer may set it according to the actual situation. The value of .

[0093] ratio Reflecting on Based on this benchmark, the assessment of signal strength for optimized routes becomes more reliable. A larger second optimization characteristic value for the optimized route indicates a smaller deviation from the corresponding initial route, resulting in richer power grid video information acquired by the hydrogen-powered drone during inspections. Furthermore, a larger second optimization characteristic value for the optimized route leads to a higher average improvement in grid signal strength within the route, which is more beneficial for information transmission by the hydrogen-powered drone, and also results in a shorter corresponding route length, saving energy consumption for the same inspection task.

[0094] (4) To further obtain a more accurate optimized route between the starting point and the endpoint, a particle swarm optimization algorithm is used for further optimization: The initial number of particles in the population is set to X. In this embodiment, X is set to 30. Each randomly obtained obstacle vector is treated as a particle. The second optimized feature value of the optimized route corresponding to each particle is used as the fitness of each particle. The maximum number of iterations is set to 20. The particle swarm optimization algorithm is then used to optimize the inspection route between the starting point and the endpoint in the optimization region, and the output is the optimal route, completing the final optimization of the inspection route between the starting point and the endpoint in the optimization region. The higher the fitness, the higher the degree of route optimization. As another embodiment of this application, the implementer can set the value of X according to the actual situation.

[0095] Finally, the optimal route for each segment of the inspection route corresponding to each optimized area is obtained using the above method, thereby obtaining the optimal route between two adjacent take-off and landing points, thus completing the planning of the hydrogen-powered UAV inspection route in the entire inspection mission.

[0096] A schematic diagram illustrating the process of obtaining RSSI correction values ​​is shown below. Figure 2 As shown.

[0097] Based on the same inventive concept as the above methods, this application also provides an inspection route planning system for hydrogen-powered drones, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described inspection route planning methods for hydrogen-powered drones.

[0098] In summary, this application provides a method for planning inspection routes for hydrogen-powered drones. It calculates the RSSI values ​​at each location within the inspection area using a signal attenuation model, serving as simulated RSSI values. A tensor constructed from these simulated values ​​is used as a benchmark, and calibration is performed using actual RSSI measurements at each location along the inspection route during each inspection. This dynamic updating during each inspection more accurately reflects the distribution of wireless communication signal strength within the inspection area. The calculation of stable RSSI distribution characteristics improves the accuracy of power distribution network cellular network evaluation, providing more reliable data support for inspection route planning. During the planning of the hydrogen-powered drone inspection route, a first optimization feature value is constructed by comprehensively considering signal strength and the validity of inspection data. When optimizing the route in segments, the size of the optimization area can be adaptively adjusted, allowing for more precise adaptation to changes in signal strength within the area. The size of the optimization area is determined by the first optimization feature value and the inspection route length. For the routes within the optimization area that require optimization, A... The inspection route is optimized using algorithms and particle swarm optimization to obtain the optimal route. Inspection is then carried out according to the optimal route, which effectively improves the inspection performance of the hydrogen-powered drone. This allows it to better adapt to changes in cellular communication signal strength during the inspection process, ensuring the stability and efficiency of data transmission, while reducing energy consumption, extending flight time, and improving overall inspection efficiency and quality.

[0099] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this application. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0100] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0101] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for planning a patrol route for a hydrogen-powered unmanned aerial vehicle, characterized in that, The method comprises the following steps: acquiring a first inspection route and each landing point on the route when a UAV first inspects an inspection area; constructing a three-dimensional grid area between adjacent two landing points, combining a signal attenuation model and a signal transmission frequency of a base station in the inspection area, calculating an RSSI simulation value of each grid in each grid area, and constructing a three-dimensional RSSI simulation tensor of each grid area; a grid on the inspection route is recorded as a route grid; based on RSSI measured values of all route grids in each grid area in all historical inspections, the RSSI simulation tensor is corrected to obtain an RSSI correction tensor of each grid area after the current inspection and an RSSI correction value of each grid; for each grid area, based on the similarity between RSSI correction tensors of the grid area after all historical adjacent two inspections, a RSSI distribution stability eigenvalue of the grid area after the current inspection is calculated; based on RSSI correction value differences and position differences between each route grid in each grid area and grids in the corresponding neighborhood during the first inspection, combining the RSSI distribution stability eigenvalue, a first optimization eigenvalue of each grid area of each inspection is constructed; for each grid area of each inspection, based on the number of route grids in the grid area and the first optimization eigenvalue, the inspection route in the grid area is segmented and the optimization area corresponding to each segment of the inspection route is determined; based on the RSSI correction value distribution characteristics of the grids in each optimization area, combining an optimization algorithm, each segment of the inspection route is optimized.

2. The method of claim 1, wherein, The acquisition process of the three-dimensional RSSI simulation tensor of each grid area is: adjacent two landing points are taken as two opposite diagonal vertices of a cuboid grid area to construct a three-dimensional grid area between the adjacent two landing points; the positions and signal transmission powers of all base stations in the inspection area are taken as inputs of the signal attenuation model to calculate the RSSI value at each grid in each grid area, which is recorded as the RSSI simulation value of each grid; a three-order tensor composed of the RSSI simulation values of all grids in each grid area is recorded as the three-dimensional RSSI simulation tensor of each grid area.

3. The method of claim 1, wherein, The acquisition process of the RSSI correction tensor of each grid area after the current inspection and the RSSI correction value of each grid is: the RSSI simulation tensor of each grid area and the RSSI measured values of all route grids in each grid area during the first inspection are taken as inputs of a Bayesian calibration method, and the output is a corrected RSSI tensor of each grid area after the first inspection, which is recorded as an RSSI correction tensor; for each grid area, the RSSI measured values of all route grids in the grid area during the current inspection and the RSSI correction tensor of the grid area after the previous inspection are taken as inputs of the Bayesian calibration method, and the output is the RSSI correction tensor of the grid area after the current inspection; each element in the RSSI correction tensor of the grid area after the current inspection is recorded as the RSSI correction value of each grid in the grid area after the current inspection.

4. The method of claim 1, wherein, The acquisition process of the RSSI distribution stability eigenvalue is: For each grid area, the similarity between the RSSI correction tensor of the grid area after each patrol and its previous patrol is recorded as the first similarity of the grid area of each patrol; the RSSI distribution stable eigenvalue of the current grid area of the current patrol is recorded as , The expression is: wherein, represents the preset selected number of inspections before the current inspection; represents the first similarity of the current grid area of the n-th inspection before the current inspection; is a normalization function.

5. The method of claim 1, wherein, The acquisition process of the first optimization eigenvalue is: In the first time of the inspection, in each grid area, the grid whose RSSI correction value in the neighborhood of any route grid is greater than the RSSI correction value of the any route grid is regarded as a strong signal grid in the neighborhood of the any route grid; calculating a first optimized feature value of a current grid region of a current patrol , The expression is: wherein, represents the RSSI distribution stability eigenvalue of the current grid area of the current patrol; represents the number of route grids in the current grid area of the first patrol; represents the number of strong signal grids in the i-th route grid neighborhood of the current grid area of the first patrol; R represents the preset neighborhood radius of the route grid; represents the distance between the i-th route grid in the current grid area of the first patrol and the j-th strong signal grid in its neighborhood; represents the RSSI correction value of the i-th route grid in the current grid area of the first patrol; represents the difference between the RSSI correction value of the i-th route grid in the current grid area of the first patrol and the j-th strong signal grid in its neighborhood; is a normalization function.

6. The method of claim 1, wherein, The step of segmenting the inspection route in the grid area and determining the optimized area corresponding to each segment of the inspection route comprises: In the current grid area of the current inspection, the product of the number of route grids and the first optimized characteristic value is calculated, and the product is denoted as a first product; a first optimized area is constructed with the starting point of the inspection route in the grid area as the center and the radius being the upward rounding result of the first product; The route grid farthest from the starting point in the inspection route after the starting point in the first optimized area is taken as the ending point in the first optimized area, and the inspection route between the starting point and the ending point in the first optimized area is obtained; The ending point in the first optimized area is taken as the next starting point, and the next optimized area and the ending point thereof are constructed by using the same obtaining method as that of the first optimized area and the ending point thereof, and the inspection route between the starting point and the ending point in the next optimized area is obtained; The construction of the new optimized area is stopped until the ending point of the optimized area is the ending point of the inspection route in the current grid area of the current inspection; and each segment of the inspection route is the inspection route between the starting point and the ending point in each optimized area.

7. The method of claim 1, wherein, The step of optimizing the route of each segment of the inspection route comprises: For any optimized area, the mean value of the RSSI correction values of all route grids in the optimized area is calculated, and the mean value is denoted as a first mean value; the grid in the optimized area smaller than the first mean value is marked as a selected obstacle grid, and all selected obstacle grids in the optimized area are sequentially sorted according to three-dimensional coordinates, and the serial number of each selected obstacle grid in the optimized area is obtained according to the sorting result; A vector composed of serial numbers of a preset number of candidate obstacle grids randomly selected from the optimization area in each time is referred to as an obstacle vector selected in each time, and all obstacle grids corresponding to the obstacle vector and the start point and the end point in the optimization area are taken as A The input of the algorithm, the output of the optimized route as the optimized route corresponding to the obstacle vector; The optimized route corresponding to any selected obstacle vector is denoted as a current optimized route, the difference between the current optimized route and the corresponding initial inspection route is analyzed, and the second optimized characteristic value of the current optimized route is constructed; For each particle in the particle swarm optimization algorithm, the obstacle vector randomly obtained each time is taken as a particle; the second optimized characteristic value of the optimized route corresponding to each particle is taken as the fitness of each particle, the particle swarm optimization algorithm is used to optimize the inspection route between the starting point and the ending point in the optimized area, and the optimal route is output.

8. The method of claim 7, wherein, The obtaining process of the second optimized characteristic value comprises: The initial route corresponding to each current optimized route is obtained; the second optimized characteristic value S of the current optimized route is calculated, and the expression of S is: wherein, represents the absolute value of the difference between the number of grids on the current optimized route and the corresponding initial route; represents the number of grids of the optimized route; represents the RSSI correction value at the mth grid on the current optimized route; represents the RSSI correction value of the grid on the initial route corresponding to the current optimized route, which is closest to the mth grid coordinate on the current optimized route; represents the distance between the mth grid on the current optimized route and the grid on the corresponding initial route, which is closest to the mth grid on the current optimized route; is a preset minimum positive number; represents an exponential function with a natural constant as the base number.

9. The method of claim 8, wherein, The obtaining process of the initial route corresponding to each current optimized route comprises: The point closest to the starting point and the ending point of the current optimized route in the first inspection route is taken as a starting point corresponding point and an ending point corresponding point respectively, and the route between the starting point corresponding point and the ending point corresponding point in the first inspection route is taken as the initial route corresponding to the current optimized route.

10. A route planning system for hydrogen-powered unmanned aerial vehicles, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor implements the steps of the method according to any one of claims 1-9 when executing the computer program.

Images