A method and system for planning the entire process of highway multi-UAV inspection

Abstract

The present invention relates to the technical field of unmanned aerial vehicle path planning, and discloses a method and system for planning the path of a highway multi-unmanned aerial vehicle inspection throughout the process, wherein the method comprises: constructing a static obstacle grid map; in the static obstacle grid map, according to the algorithm of the adaptive expansion field, respectively determining the first path and the second path of the unmanned aerial vehicle; according to the first path and the second path, respectively calculating the first path length and the second path length, and judging whether the path length of the inspection throughout the process meets the cruising distance requirement of the unmanned aerial vehicle according to the first path length and the second path length; according to the driving radius of the unmanned aerial vehicle and the width of the road section to be inspected, determining whether the road section to be inspected can be covered for inspection; determining the third path, determining the fourth path, determining the fifth path, and combining the third, fourth and fifth paths of each unmanned aerial vehicle to obtain the complete path of each unmanned aerial vehicle. The present invention is convenient to use, and the planning results are comprehensive and practical.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) path planning, and in particular to a method and system for planning the entire process of highway multi-UAV inspection paths. Background Art

[0002] With the continuous increase in the mileage of highways, the demand for efficient maintenance and management of highways is becoming more and more urgent. However, traditional manual inspections and fixed cameras based on video stream recognition are inefficient and costly, and are increasingly unable to meet the demand. In addition, since manual inspections are easily affected by traffic flow, it is difficult to ensure the coverage of the inspections, and the inspection method of fixed cameras is plagued by problems such as uneven equipment and detection blind spots, and it is almost impossible to further optimize on the existing basis. Drones have the advantages of not being affected by ground traffic, strong mobility and low maintenance costs. The method of using multiple drones equipped with various sensors to inspect highways has attracted widespread attention.

[0003] The core problem that needs to be solved in order to realize highway inspection by multiple drones is how to plan their paths. Most of the existing multi-drone path planning methods are extensions of the "traveling salesman" problem, that is, treating multiple drones as a particle and planning the optimal path for them from a macro perspective, ignoring the path planning of multiple drones in each section of the inspection. Methods such as deep learning and reinforcement learning are also common, but they require a large amount of original data collection and training, which are prone to local failures caused by "overfitting" and other problems. In addition, they mainly solve the problem of facing static obstacles. For dynamic obstacles that suddenly appear during the inspection process, they can only rely on the drone's own control system to avoid them, and cannot make advance predictions. For the path planning of multiple drones performing inspection tasks, the inspection purpose is mainly achieved by dividing the inspection area. This method is incapable of being directly applied when there is a section in a certain area of ​​the highway inspection section that a single drone cannot pass through. In addition, the existing multi-drone path planning method research rarely focuses on highways, especially the path planning of the entire process from the drone take-off point to the inspection section and then back to the take-off point is almost a current research gap. Therefore, how to perform reasonable and efficient full-process path planning for multiple UAVs during each highway inspection process has become a key issue that needs to be solved urgently.

[0004] Swarm intelligence refers to the complex behaviors and patterns generated by multiple individuals through group rules and interactions. It has advantages such as strong adaptability and superior self-organization, but it is prone to unavoidable shortcomings such as poor stability and low search efficiency when solving multi-UAV path planning problems. Although traditional intelligent algorithms such as algorithms have the advantages of superior robustness and accurate calculation, they are difficult to complete path planning tasks independently because the results of their calculations can only get one path, can only solve part of the problem in a fixed area, and have low search efficiency. Summary of the invention

[0005] In order to address the deficiencies in the prior art, the present invention provides a method and system for full-process path planning for multi-UAV inspection of highways.

[0006] On the one hand, a method for planning the whole process of highway multi-UAV inspection path is provided, including: constructing a static obstacle grid map; in the static obstacle grid map, according to the adaptive expansion field The algorithm determines the first path and the second path of the drone respectively; according to the first path and the second path, the first path length and the second path length are calculated respectively; according to the first path length and the second path length, whether the path length of the whole inspection process meets the cruising distance requirement of the drone is judged; if it meets, the number of drones input by the user is obtained, if it does not meet, a prompt is given; according to the driving radius of the drone and the width of the section to be inspected, whether the section to be inspected can be covered for inspection is determined; if it cannot be covered for inspection, the number of drones input by the user is obtained again; if it can be covered for inspection, the take-off point of the drone is determined; the third path is determined, and the third path is the refined path between the take-off point of each drone at the time of departure and the inspection starting point at the starting end of the section to be inspected; the fourth path is determined, and the fourth path is the refined path between the inspection starting point of each drone at the starting end of the section to be inspected and the midpoint of the terminal end of the section to be inspected; the fifth path is determined, and the fifth path is the return path of each drone from the midpoint of the terminal end of the section to be inspected to the starting point; the third, fourth and fifth paths of each drone are combined to obtain the complete path of each drone.

[0007] On the other hand, a highway multi-UAV inspection full-process path planning system is provided, including: a map construction module, which is configured to: construct a static obstacle grid map; within the static obstacle grid map, according to the adaptive expansion field The algorithm determines the first path and the second path of the drone respectively; the judgment module is configured to: calculate the first path length and the second path length respectively according to the first path and the second path, and judge whether the path length of the whole inspection process meets the cruising distance requirement of the drone according to the first path length and the second path length; if it meets, the number of drones input by the user is obtained, and if it does not meet, a prompt is given; the determination module is configured to: determine whether the road section to be inspected can be covered and inspected according to the driving radius of the drone and the width of the road section to be inspected; if it cannot be covered and inspected, the number of drones input by the user is obtained again; if it can If the inspection is covered, the take-off point of the UAV is determined; a planning module is configured to: determine a third path, which is a refined path for each UAV from the take-off point at the time of departure to the inspection starting point at the starting end of the section to be inspected; determine a fourth path, which is a refined path for each UAV from the inspection starting point at the starting end of the section to be inspected to the midpoint of the terminating end of the section to be inspected; determine a fifth path, which is a return path for each UAV from the midpoint of the terminating end of the section to be inspected to the starting point; combine the third, fourth and fifth paths of each UAV to obtain a complete path for each UAV.

[0008] The above technical solution has the following advantages or beneficial effects: 1. The present invention comprehensively considers the whole process of drone inspection and divides it into two parts: the drone goes back and forth to inspect the road section. The algorithm is used in combination with a swarm intelligence method that uses a Gaussian distribution for local search, which guarantees the fuel consumption of the drone to the greatest extent and divides the inspection area into partitions, and merges areas where there are sections that cannot be passed through, ensuring the inspection coverage rate. Compared with other methods, it solves the problem of only considering macroscopic paths and failing to guarantee the drone path during the inspection process, and also avoids the pain point of requiring a large amount of data support for training to effectively plan the path. It is easy to use, and the planning results are comprehensive and practical. 2. The present invention performs chaotic planning on the take-off points of the drones so that they are evenly distributed within a certain range, and performs local search through Gaussian distribution. At the same time, the step size is continuously reduced during the continuous evolution of the group, which greatly reduces the risk of falling into a local optimal solution, and gives full play to the superior self-organizing characteristics of the swarm intelligence method, and outputs the path of each drone separately, fully ensuring the feasibility of this method. 3. The present invention improves the traditional The algorithm is improved so that it only needs to traverse the adjacent grids between the current grid and the target grid for expansion, which greatly improves the calculation efficiency and guarantees the calculation results. The algorithm performs initial path planning, checks whether the drone's cruising length can meet the requirements, avoids invalid calculations, and provides a map basis through plane transformation during the inspection process to ensure inspection coverage. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0010] Figure 1 This is a flow chart of the method of embodiment 1.

[0011] Figure 2 This is the route obstacle avoidance method of embodiment 1. DETAILED DESCRIPTION

[0012] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0013] Terminology: Chaos algorithm is a nonlinear dynamic system with high sensitivity and unpredictability. Chaos is a general term for random behavior exhibited by deterministic systems, and its root lies in nonlinear interactions. A small change in the initial conditions of a chaotic system will lead to a huge change in the system output.

[0014] The Dubins curve consists of three basic movements: left turn (L), straight (S), and right turn (R). The combination of these three basic movements constitutes six different Dubins curves, namely LSL, RSR, RSL, LSR, RLR, and LRL. Each curve represents the optimal path from the starting point to the end point, and the number of turns and angles involved are optimized. It describes the optimal path curve formed by minimizing the driving distance or time under the given starting point and end point.

[0015] Gaussian distribution is a continuous probability distribution for random variables, which is symmetrical. Its mathematical expectation is Determines the location, variance determines the magnitude of the distribution. The standard normal distribution is a special case where and .

[0016] Embodiment 1: Figure 1 As shown, this embodiment provides a method for planning the path of a highway multi-UAV inspection process, including: S101: constructing a static obstacle grid map; in the static obstacle grid map, according to the adaptive expansion field The algorithm determines the first path and the second path of the UAV respectively.

[0017] S102: Calculate the first path length and the second path length respectively according to the first path and the second path length, and judge whether the path length of the whole inspection process meets the cruising distance requirement of the UAV according to the first path length and the second path length; if it meets the requirement, enter S103; if it does not meet the requirement, give a prompt.

[0018] S103: Obtain the number of drones input by the user.

[0019] S104: Determine whether the road section to be inspected can be covered and inspected according to the driving radius of the drone and the width of the road section to be inspected; if it cannot be covered and inspected, return to S103; if it can be covered and inspected, determine the take-off point when the drone departs.

[0020] S105: Determine a third path, which is a refined path of each UAV from the take-off point at the time of departure to the inspection starting point at the starting end of the section to be inspected; determine a fourth path, which is a refined path of each UAV from the inspection starting point at the starting end of the section to be inspected to the midpoint of the terminating end of the section to be inspected; determine a fifth path, which is a return path of each UAV from the midpoint of the terminating end of the section to be inspected to the starting point; combine the third, fourth and fifth paths of each UAV to obtain a complete path for each UAV.

[0021] Furthermore, the first path refers to: a rough path from the starting point of the drone to the center point of the starting end of the section to be inspected; the second path refers to: a rough path from the center point of the starting end of the section to be inspected to the center point of the ending end of the section to be inspected.

[0022] Furthermore, the method also includes: S106: determining whether there is a path conflict between the complete paths of the UAVs, if not, directly outputting the path, if yes, fine-tuning the path.

[0023] Furthermore, the method also includes: S107: detecting dynamic obstacles through a highway gantry camera to determine whether re-path planning is required, and if so, returning to the step of determining the third path.

[0024] Furthermore, the S101: static obstacle grid map is obtained by vertically projecting the static obstacles in the three areas, and the vertical projection points of the static obstacles in the three areas are used as the static obstacle taboo point set; the three areas are respectively: the area from the take-off point of the drone to the center point of the starting end of the section to be inspected, the area from the center point of the starting end of the section to be inspected to the center point of the ending end of the section to be inspected, and the area from the center point of the ending end of the section to be inspected to the starting point of the drone. The distance from the ending end of the section to be inspected to the starting point is greater than the distance from the starting end of the section to be inspected to the starting point.

[0025] It should be understood that a static obstacle grid map is established, including the vertical projection of obstacles in the section and in the flight area of ​​all drones from the starting point of the drone to the center point of the starting end of the section to be inspected, to obtain the taboo point set , set the horizontal center point of the far end section of the road segment as the fourth path planning end point , set the number of iterations.

[0026] Further, the step S101: in the static obstacle grid map, according to the adaptive expansion field The algorithm determines the first path and the second path of the UAV respectively; wherein the adaptive expansion field The algorithm specifically includes: S101-1: determining the traversal range of the grid according to the grid coordinates of the starting point and the coordinates of the center point of the starting end of the section to be inspected; S101-2: calculating the cost of each grid within the traversal range of the grid; S101-3: sorting the costs of the grids in descending order to determine the extended grid; S101-4: if the extended grid is the coordinates of the center point of the starting end of the section to be inspected, then ending and obtaining the first path.

[0027] Among them, the steps for determining the second path are the same as those for the first path. The only difference is that the grid coordinates of the starting point of S101-1 are replaced by the coordinates of the center point of the starting end of the section to be inspected, and the coordinates of the center point of the starting end of the section to be inspected of S101-1 are replaced by the coordinates of the center point of the ending end of the section to be inspected.

[0028] Exemplarily, the traversal range of the grid is determined according to the grid coordinates of the starting point and the coordinates of the center point of the starting end of the road section to be inspected, and only the grids whose horizontal coordinates and vertical coordinates are within the horizontal coordinates and vertical coordinates of the grid of the center point of the starting end of the road section to be inspected are selected; for example, the coordinates of the grid of the starting point are , the coordinates of the center point of the starting end of the road section to be inspected are , then the traversal range of the grid is the horizontal coordinate in In the range, the vertical axis is The grid within the range.

[0029] Furthermore, the S101-2: calculating the cost of each grid within the traversal range of the grid includes: ;in, is the cumulative cost from the starting grid to the calculation grid, is the sum of the distances between adjacent points included in the path after each search; is a heuristic value, and it is recommended to be the straight-line distance from the calculation grid to the destination; For Grid The cost.

[0030] Furthermore, the S101-3: sorting the costs of the grids in descending order to determine the extended grid specifically includes: the extended grid is the grid with the smallest cost among all search grids, that is, the grid included in the path among all search grids.

[0031] Furthermore, the S102: according to the first path and the second path, respectively calculate the first path length and the second path length, and the specific calculation process includes: for the first path, calculating the path length according to the Dubins curve to obtain the first path length; for the second path, calculating the path length according to the Dubins curve to obtain the second path length.

[0032] Further, the S102: judging whether the path length of the entire inspection process meets the cruising distance requirement of the UAV based on the first path length and the second path length, specifically includes: summing the first path length and the second path length, and comparing the sum result with the maximum cruising radius of the UAV itself. If the sum result is less than the maximum cruising radius of the UAV itself, it means that the cruising distance requirement of the UAV is met; otherwise, it means that the cruising distance requirement of the UAV is not met.

[0033] Further, S103: obtaining the number of drones determined by the user; wherein the number of drones is set according to the actual number of drones, and if the number of drones is less than four, the number of drones is set to four, and finally only the paths matching the actual number of drones are output.

[0034] Furthermore, S104: determines whether the road section to be inspected can be covered and inspected based on the driving radius of the drone and the width of the road section to be inspected, specifically including: judging whether the sum of the actual driving radii of the inspection drones is less than the width of the inspection section, if so, it indicates that it cannot be covered, and if so, it indicates that it can be covered.

[0035] Furthermore, the S104: if the inspection can be covered, then determine the take-off point of the drone, specifically including: S104-1: determine the take-off range of the drone at the starting point; S104-2: generate a random number ; represents the first random number, Represents the second random number; S104-3: Generate the take-off point when the drone departs, the calculation function is: ; ; ;in, Represents the take-off point of a drone. Respectively represent the center coordinates of the drone’s take-off range, Represents the radius of the drone's takeoff range, Represents a set of taboo points.

[0036] Further, the S105: determining a third path, the third path is a refined path between each drone from the take-off point at the departure to the inspection starting point at the starting end of the section to be inspected, including: S105-11: calculating fitness, the first three points in descending order of fitness are recorded as ; The fitness function is: ;in, For the The fitness of each drone; Representative The straight-line distance from the current position of each drone to the end position; Representative The number of obstacles between each drone and the finish line; Represents the sum of the number of obstacles between all drones and the finish line.

[0037] S105-12: Update the positions of the remaining points: .

[0038] .

[0039] .

[0040] .

[0041] .

[0042] in, From 2 to 0 according to the number of iterations, is a random number in the interval [0,1], Represents the updated coordinates of the remaining points. This updating method can ensure that as the population continues to evolve, the overall fitness function decreases, while also retaining the exploration of the solution space. represents the search range correction factor, Represents the search divergence correction coefficient, which is used to ensure the diversity of the search range and search direction.

[0043] S105-13: Perform local search based on Gaussian distribution.

[0044] ;in, for[ , ]A random number in the range, and is a dynamic factor. When unchanged, The larger the value, the smaller the local search range. Local search correction The extreme cases generated during the update and detailed exploration of the space within the existing range, while preventing falling into the local optimal solution, represents a unit vector; express Coordinate vector of the moment; express Coordinate vector of the moment; express The coordinate vector after local search at time instant.

[0045] S105-14: UPDATE location.

[0046] . . .

[0047] .

[0048] in, is the dynamic factor, The larger it is, the smaller the iteration step size is; is the current iteration number, macro-control search accuracy, with continuous iteration, As the number of nodes increases, the group gets closer to the destination and the search accuracy increases, ensuring the stability of the algorithm and preventing it from falling into a local optimal solution.

[0049] S105-15: Record the position of each point.

[0050] S105-16: Has the maximum number of iterations been reached? If so, output the path and set the end points of the path to evenly distributed points according to S105-15; if not, return to S105-11.

[0051] Furthermore, S105: determines a fourth path, which is a fine path of each UAV from the inspection starting point at the starting end of the road section to be inspected to the ending end of the road section to be inspected, specifically including: S105-21: according to the UAV inspection radius and the road width, the road is evenly divided into several UAV inspection areas along a direction parallel to the longitudinal center axis of the road.

[0052] S105-22: According to the inspection areas divided in S105-21, a total taboo point set is established for each area, and the total taboo point set includes: a static obstacle taboo point set and a dynamic obstacle taboo point set; the dynamic obstacle is a dynamic obstacle detected by the camera.

[0053] S105-23: Determine whether there is a vertical section in each inspection area that cannot pass the drone. If not, go to step S105-24; if so, go to step S105-25; the vertical section refers to the plane in each inspection area that is perpendicular to the road surface.

[0054] S105-24: The center line of the road surface in each inspection area is regarded as the length of the grid map, and the flying height of the drone is regarded as the height of the grid map. A grid map is established in the rectangular area composed of the length and height, with the midpoint of the inspection area corresponding to the starting end of the inspection section as the starting point, and the midpoint of the inspection area corresponding to the end end of the inspection section as the end point. The grid map is constructed according to the adaptive expansion field. Algorithm for path planning.

[0055] S105-25: Integrate the inspection area with the impassable section obtained in S105-23 with one or two adjacent inspection areas into one inspection area, and use the inspection drones corresponding to the areas before the merger; search the remaining areas according to step S105-24.

[0056] Furthermore, S105: determines the fifth path, which is the return path of each drone from the midpoint of the terminal end of the section to be inspected to the starting point, specifically including: setting the midpoint of the terminal end of the section to be inspected as the starting point, setting the starting point as the destination, and using the method for determining the third path to perform path planning to obtain the fifth path.

[0057] Furthermore, the S105: combining the third, fourth and fifth paths of each drone to obtain a complete path of each drone specifically includes: connecting the third, fourth and fifth paths of each drone in sequence to obtain the complete path of the current drone.

[0058] According to the above method, the driving distance of each drone can be relatively even, avoiding the risk of being unable to return.

[0059] Furthermore, the S106: determines whether there is a path conflict between the complete paths of the UAVs. If not, the path is directly outputted. If yes, the path is fine-tuned, specifically including: S106-1: calculates the moment when the conflicting UAVs arrive at the path conflict point at the set speed. If the difference in the moment when different UAVs arrive at the conflict point is greater than the set time range, go to S107; otherwise, go to S106-2; the path conflict means: if there is an intersection between the safety circles of the two UAVs, it means that there is a path conflict between the paths of the two UAVs; the UAV safety circle refers to a circle with the center point of the UAV as the center and the radius covered by the UAV during flight as the radius.

[0060] S106-2: Assume that the first UAV safety circle is A, and the second UAV safety circle is B; the center of safety circle A is O, and the center of safety circle B is P. Assume that safety circle A intersects with safety circle B to obtain intersection points C and D, connect intersection points C and D to obtain line CD, and then obtain the midpoint E of line CD, identify the number of UAVs on both sides of line CD, select the center of the side with relatively fewer UAVs, assume that the selected center is O, then connect point E and point O, and draw the extension line of line EO, the extension line EO intersects with safety circle A at point F, such as Figure 2 shown.

[0061] S106-3: Move the center O of the safety circle to the point F generated in S106-2, set the newly generated intersection F as the waypoint, and the conflict point is no longer used.

[0062] Furthermore, the S107: detecting a dynamic obstacle and determining whether re-path planning is required, and if so, returning to S101, specifically including: if the dynamic obstacle coincides with the currently planned path, re-planning is required, otherwise, re-planning is not required.

[0063] The present invention first The algorithm is improved so that it only needs to traverse the adjacent grids between the current grid and the target grid for expansion selection, and through the improved The algorithm calculates the initial path considering only static obstacles, detects the drone’s inability to return due to energy constraints, and further generates the initial take-off point through the chaos algorithm, plans the path from the starting point to the inspection section in a swarm intelligence way, and performs local search through high-speed distribution, thereby achieving improved The algorithm is combined with a swarm intelligence method that uses Gaussian distribution to perform regional search, partition inspection routes, and plan return routes, thus achieving full-process path planning for multiple inspection drones on highways.

[0064] The present invention comprehensively considers the path planning tasks at different times during the drone inspection process, provides a full-process path from departure to inspection and then to return, and detects dynamic obstacles during flight through a camera, meeting multiple requirements such as inspection coverage, fuel constraints, and obstacle avoidance. In addition, through real-time detection of dynamic obstacles, the risk of sudden changes in dynamic obstacles during the execution of the drone mission can be avoided, providing a more comprehensive and reasonable path planning solution for multi-drone path planning.

[0065] Embodiment 2: This embodiment provides a highway multi-UAV inspection full-process path planning system, including: a map construction module, which is configured to: construct a static obstacle grid map; within the static obstacle grid map, according to the adaptive expansion field The algorithm determines the first path and the second path of the UAV respectively.

[0066] The judgment module is configured to: calculate the first path length and the second path length respectively according to the first path and the second path, and judge whether the path length of the entire inspection process meets the cruising distance requirement of the drone according to the first path length and the second path length; if it meets the requirement, obtain the number of drones input by the user, and if it does not meet the requirement, give a prompt.

[0067] The determination module is configured to: determine whether the road section to be inspected can be covered and inspected based on the driving radius of the drone and the width of the road section to be inspected; if it cannot be covered and inspected, re-acquire the number of drones input by the user; if it can be covered and inspected, determine the take-off point when the drone departs.

[0068] The planning module is configured to: determine a third path, which is a refined path for each drone from its take-off point at the time of departure to the inspection starting point at the starting end of the section to be inspected; determine a fourth path, which is a refined path for each drone from the inspection starting point at the starting end of the section to be inspected to the midpoint of the terminating end of the section to be inspected; determine a fifth path, which is a return path for each drone from the midpoint of the terminating end of the section to be inspected to the starting point; and combine the third, fourth and fifth paths of each drone to obtain a complete path for each drone.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for planning the entire process of highway multi-UAV inspection path, characterized in that: include: Construct a static obstacle grid map; In the static obstacle grid map, according to the adaptive expansion of the field The algorithm determines the first path and the second path of the drone respectively; calculates the first path length and the second path length respectively according to the first path and the second path; and determines whether the path length of the entire inspection process meets the cruising distance requirement of the drone according to the first path length and the second path length; If it matches, the number of drones entered by the user is obtained; if it does not match, a prompt is given; Among them, the adaptive expansion field Algorithms, including: Determine the traversal range of the grid according to the grid coordinates of the starting point and the coordinates of the center point of the starting end of the road section to be inspected; Calculate the cost of each grid within the traversal range of the grid; Sort the grid costs from high to low to determine the expanded grid; If the extended grid is the coordinate of the center point of the starting end of the road section to be inspected, the process ends and the first path is obtained; According to the driving radius of the drone and the width of the road section to be inspected, determine whether the road section to be inspected can be covered for inspection; if it cannot be covered for inspection, re-acquire the number of drones input by the user; if it can be covered for inspection, determine the take-off point for the drone to depart, including: Determine the takeoff range of the drone at the departure point; Generate random numbers ; represents the first random number, represents the second random number; Generate the take-off point when the drone departs, the calculation function is: ； ； ； in, Represents the take-off point of a drone. Respectively represent the center coordinates of the drone’s take-off range, Represents the radius of the drone's takeoff range, represents a set of taboo points; Determine a third path, which is the refined path of each UAV from the take-off point at the time of departure to the inspection starting point at the starting end of the section to be inspected; determine a fourth path, which is the refined path of each UAV from the inspection starting point at the starting end of the section to be inspected to the terminating end of the section to be inspected; determine a fifth path, which is the return path of each UAV from the midpoint of the terminating end of the section to be inspected to the starting point; combine the third, fourth and fifth paths of each UAV to obtain a complete path for each UAV.

2. A method for planning the entire process of highway multi-UAV inspection path as claimed in claim 1, characterized in that: The first path refers to: a rough path of the drone from the starting point to the center point of the starting end of the section to be inspected; the second path refers to: a rough path from the center point of the starting end of the section to be inspected to the center point of the ending end of the section to be inspected; The static obstacle grid map is obtained by vertically projecting the static obstacles in three areas, and the vertical projection points of the static obstacles in the three areas are used as the static obstacle taboo point set; the three areas are: the area between the take-off point of the drone and the center point of the starting end of the section to be inspected, the area between the center point of the starting end of the section to be inspected and the center point of the ending end of the section to be inspected, and the area between the center point of the ending end of the section to be inspected and the starting point of the drone; The distance from the end of the road section to be inspected to the starting point is greater than the distance from the start of the road section to be inspected to the starting point.

3. A method for planning the entire process of highway multi-UAV inspection path as claimed in claim 1, characterized in that: According to the first path length and the second path length, it is judged whether the path length of the whole inspection process meets the cruising distance requirements of the UAV, specifically including: The first path length and the second path length are summed, and the sum result is compared with the maximum cruising radius of the UAV itself. If the sum result is smaller than the maximum cruising radius of the UAV itself, it means that the cruising distance requirement of the UAV is met; otherwise, it means that the cruising distance requirement of the UAV is not met.

4. A method for planning the entire process of highway multi-UAV inspection path as claimed in claim 1, characterized in that: Determine a third path, which is a refined path between each drone's take-off point at the time of departure and the inspection starting point at the starting end of the section to be inspected, specifically including: Step (1-1): Calculate the fitness. The first three points of fitness from small to large are recorded as ; The fitness function is: ； in, For the The fitness of each drone; Representative The straight-line distance from the current position of each drone to the end position; Representative The number of obstacles between each drone and the finish line; Represents the sum of the number of obstacles between all drones and the finish line; Step (1-2): Update the positions of the remaining points: ； ； ； ； ； in, From 2 to 0 according to the number of iterations, is a random number in the interval [0,1], Represents the updated coordinates of the remaining points, represents the search range correction factor, represents the search divergence correction factor; Steps (1-3): Local search based on Gaussian distribution: ； in, for[ , ]A random number in the range, and is a dynamic factor; When unchanged, The larger the local search range, the smaller the local search range; local search correction The extreme cases generated during the update and detailed exploration of the space within the existing range, while preventing falling into the local optimal solution, represents a unit vector; express Coordinate vector of the moment; express Coordinate vector of the moment; express The coordinate vector after local search at time instant; Steps (1-4): Update Location: ； ； ； ； in, is the dynamic factor, The larger it is, the smaller the iteration step size is; is the current iteration number, macro-control search accuracy, with continuous iteration, As the number of nodes increases, the group gets closer to the destination, and the search accuracy gradually increases, ensuring the stability of the algorithm and preventing it from falling into a local optimal solution; Steps (1-5): Record the position of each point; Step (1-6): Has the maximum number of iterations been reached? If so, output the path and set the end points of the path to evenly distributed points according to step (1-5); if not, return to step (1-1).

5. A method for planning the entire process of highway multi-UAV inspection path as claimed in claim 1, characterized in that: Determine a fourth path, where the fourth path is a refined path between the inspection starting point of each drone at the starting end of the road section to be inspected and the ending end of the road section to be inspected, specifically including: Step (2-1): According to the inspection radius of the drone and the width of the road, the road is evenly divided into a number of drone inspection areas along a direction parallel to the longitudinal center axis of the road; Step (2-2): according to the inspection areas divided in step (2-1), a total taboo point set is established for each area, wherein the total taboo point set includes: a static obstacle taboo point set and a dynamic obstacle taboo point set; The dynamic obstacle is a dynamic obstacle detected by a camera; Step (2-3): Determine whether there is a vertical section in each inspection area that cannot be passed by the drone. If not, go to step (2-4); if so, go to step (2-5); the vertical section refers to the plane perpendicular to the road surface in each inspection area; Step (2-4): Consider the center line of the road surface in each inspection area as the length of the grid map, and the flight height of the drone as the height of the grid map. Create a grid map in the rectangular area composed of the length and height, starting from the midpoint of the inspection area corresponding to the starting end of the inspection section and ending at the midpoint of the inspection area corresponding to the end of the inspection section. Algorithms for path planning; Step (2-5): The inspection area with the impassable section obtained in step (2-3) is integrated with one or two adjacent inspection areas into one inspection area, and the inspection drones are the drones corresponding to the areas before the merger; the remaining areas are searched according to step (2-4).

6. A method for planning the path of a highway multi-UAV inspection process as claimed in claim 1, characterized in that: The method further comprises: Determine whether there is a path conflict between the complete paths of the drones. If not, directly output the path. If yes, fine-tune the path. Determine whether there is a path conflict between the complete paths of the drones. If not, directly output the path. If yes, fine-tune the path, including: Step (3-1): Calculate the time when the conflicting UAVs arrive at the path conflict point at the set speed. If the difference in the time when different UAVs arrive at the conflict point is greater than the set time range, go to the step of determining whether the path needs to be re-planned. If not, go to step (3-2); the path conflict means: if there is an intersection between the safety circles of the two UAVs, it means that there is a path conflict between the paths of the two UAVs; the UAV safety circle refers to a circle with the center point of the UAV as the center and the radius covered by the UAV during flight as the radius; Step (3-2): Assume that the first drone safety circle is A, and the second drone safety circle is B; the center of safety circle A is O, and the center of safety circle B is P. Assume that safety circle A intersects with safety circle B to obtain intersection points C and D, connect intersection points C and D to obtain line CD, and then obtain the midpoint E of line CD, identify the number of drones on both sides of line CD, select the center of the side with relatively fewer drones, and assume that the selected center is O, then connect point E and point O, and draw the extension line of line EO, and the extension line EO intersects with safety circle A at point F; Step (3-3): Move the center O of the safety circle to point F generated in step (3-2), set the newly generated intersection point F as the approach point, and the conflict point is no longer used.

7. A method for planning the entire process of highway multi-UAV inspection path as claimed in claim 1, characterized in that: The method further comprises: Dynamic obstacles are detected by the highway gantry camera to determine whether re-path planning is required. If so, return to the third path determination step; if the dynamic obstacle coincides with the current planned path, re-planning is required, otherwise, no re-planning is required.

8. A highway multi-UAV inspection full-process path planning system, characterized by: include: A map construction module, which is configured to: construct a static obstacle grid map; In the static obstacle grid map, according to the adaptive expansion of the field Algorithms,determine the first path and the second path of the UAV respectively; Among them, the adaptive expansion field Algorithms, including: Determine the traversal range of the grid according to the grid coordinates of the starting point and the coordinates of the center point of the starting end of the road section to be inspected; Calculate the cost of each grid within the traversal range of the grid; Sort the grid costs from high to low to determine the expanded grid; If the extended grid is the coordinate of the center point of the starting end of the road section to be inspected, the process ends and the first path is obtained; The judgment module is configured to: calculate the first path length and the second path length respectively according to the first path and the second path, and judge whether the path length of the whole inspection process meets the cruising distance requirement of the drone according to the first path length and the second path length; if it meets, obtain the number of drones input by the user, and if it does not meet, give a prompt; The determination module is configured to: determine whether the road section to be inspected can be covered and inspected according to the driving radius of the drone and the width of the road section to be inspected; if it cannot be covered and inspected, re-acquire the number of drones input by the user; if it can be covered and inspected, determine the take-off point when the drone departs, specifically including: Determine the takeoff range of the drone at the departure point; Generate random numbers ; represents the first random number, represents the second random number; Generate the take-off point when the drone departs, the calculation function is: ； ； ； in, Represents the take-off point of a drone. Respectively represent the center coordinates of the drone’s take-off range, Represents the radius of the drone's takeoff range, represents a set of taboo points; The planning module is configured to: determine a third path, which is a refined path for each drone from its take-off point at the time of departure to the inspection starting point at the starting end of the section to be inspected; determine a fourth path, which is a refined path for each drone from the inspection starting point at the starting end of the section to be inspected to the midpoint of the terminating end of the section to be inspected; determine a fifth path, which is a return path for each drone from the midpoint of the terminating end of the section to be inspected to the starting point; and combine the third, fourth and fifth paths of each drone to obtain a complete path for each drone.

Images