A method for collaborative three-dimensional inspection planning of heterogeneous robots in vertical structure facilities

Abstract

The present invention relates to the field of robotics technology, and in particular to a method for collaborative three-dimensional inspection planning of heterogeneous robots under vertical structure facilities. The point cloud data of the target facility of the vertical structure is obtained; the minimum convex polygon describing the target facility of the vertical structure is generated by a three-dimensional convex hull enhancement algorithm based on the point cloud data; the cruising start point and cruising end point of the UAV are determined based on the minimum convex polygon; the improved A-star algorithm is used to perform path planning between the cruising start point and the cruising end point of the UAV to determine the global inspection path of the UAV; an inspection is performed according to the determined global inspection path to obtain a global inspection result; the position to be inspected of the climbing robot in the minimum convex polygon is determined based on the global inspection result; the improved branch and bound algorithm is used to determine the vertical inspection path of the climbing robot based on the position to be inspected. The present invention can achieve more efficient, safer and more intelligent inspection operations, while covering a wider area and improving the flexibility and efficiency of task execution.

Description

Technical Field

[0001] The present invention relates to the field of robotics technology, and in particular to a method for planning a coordinated three-dimensional inspection of heterogeneous robots in a vertical structure facility. Background Art

[0002] Facility inspection plays an important role in maintaining the functionality of facilities. In early applications of robotic inspection, robot-based methods provide higher efficiency and flexibility than manual solutions. Robotic systems can get closer to the target and obtain information at a given location with higher accuracy, but their field of view is limited. In addition, robots are prone to encounter obstacles during the inspection process, causing the inspection to fall into local optimality. Later, researchers explored the use of drones for facility-related inspections to solve this problem. Drones are more flexible and have a better field of view than robots. They can quickly cover large areas of facilities, but their payloads are smaller and the operation time is shorter. However, they are susceptible to flight time and interference from the external environment, and cannot achieve contact inspections. In recent years, there has been more and more research on heterogeneous robotic systems. These two types of equipment can complement each other and improve the overall performance of detection, but for heterogeneous systems, the huge differences in sensor settings, maneuverability, perspectives, and processing capabilities pose challenges to coordination strategies.

[0003] To address this challenge, and in order to solve the problem of insufficient coordination and processing capabilities during the inspection process of heterogeneous robot systems, the present invention proposes a collaborative three-dimensional inspection path planning method for land and air heterogeneous robots under vertical structure facilities. Summary of the Invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a method for collaborative three-dimensional inspection planning of heterogeneous robots under vertical structure facilities. The method obtains point cloud data of the target facility with a vertical structure; generates a minimum convex polygon describing the target facility with a vertical structure based on the point cloud data using a three-dimensional convex hull enhancement algorithm; determines the cruise start and cruise end of the drone based on the minimum convex polygon; uses an improved A-star algorithm to perform path planning between the cruise start and cruise end of the drone to determine the drone's global inspection path; performs inspection according to the determined global inspection path to obtain a global inspection result; determines the position to be inspected of the climbing robot in the minimum convex polygon based on the global inspection result; and uses an improved branch and bound algorithm to determine the vertical inspection path of the climbing robot based on the position to be inspected. The present invention can achieve more efficient, safer, and more intelligent inspection operations, while covering a wider area and improving the flexibility and efficiency of task execution.

[0005] The present invention adopts the following technical solution, a method for planning a coordinated three-dimensional inspection of heterogeneous robots in a vertical structure facility, comprising:

[0006] Acquire point cloud data of a target facility with a vertical structure; and generate a minimum convex polygon describing the target facility with a vertical structure using a three-dimensional convex hull enhancement algorithm based on the point cloud data;

[0007] Determine a cruising start point and a cruising end point of the UAV based on the minimum convex polygon;

[0008] The improved A-star algorithm is used to plan the path between the UAV's cruise start point and cruise end point to determine the UAV's global inspection path;

[0009] The drone performs inspection according to the determined global inspection path and obtains a global inspection result;

[0010] Determining a position to be inspected of the climbing robot in the minimum convex polygon according to the global inspection result;

[0011] According to the position to be inspected, an improved branch and bound algorithm is adopted to determine the vertical inspection path of the climbing robot.

[0012] Furthermore, a three-dimensional convex hull enhancement algorithm is used based on the point cloud data to generate a minimum convex polygon of the target facility describing the vertical structure, specifically:

[0013] Select three non-collinear point clouds in the target facility point cloud data of the vertical structure as three vertices to establish the initial convex hull;

[0014] Select any point cloud in the target facility point cloud data of the vertical structure and determine whether the point cloud is inside the initial convex hull. If it is inside, skip it.

[0015] If it is not inside, determine the positional relationship between the point cloud and each edge of the initial convex hull, and add the point cloud to the initial convex hull based on the positional relationship;

[0016] All point clouds in the target facility point cloud data of the vertical structure are traversed in sequence to obtain a convex hull point cloud set of the target facility that describes the vertical structure;

[0017] The minimum convex polygon of the target facility describing the vertical structure is obtained based on the convex hull point cloud set.

[0018] Furthermore, the improved A-star algorithm is used to plan the path between the cruise start point and the cruise end point of the UAV, including:

[0019] Setting a heuristic function, wherein the heuristic function is used to estimate the cost of the UAV from the cruise start point to the cruise end point;

[0020] Construct a cost function based on the turning cost, rising cost and collision avoidance cost of the UAV from the cruising start point to the cruising end point;

[0021] Searching for a next viewpoint from the cruise starting point, calculating the priority of the next viewpoint according to the heuristic function and the cost function, and selecting the next viewpoint with the highest priority as the optimal viewpoint;

[0022] Taking the optimal viewpoint as the starting point, search for the next viewpoint again and select the next viewpoint with the highest priority as the optimal viewpoint;

[0023] Iterate in sequence until the cruise end is reached and all optimal viewpoints are obtained;

[0024] The global inspection path of the UAV is obtained based on all the optimal viewpoints.

[0025] Furthermore, when obtaining the global inspection path of the UAV based on all the optimal viewpoints, it also includes:

[0026] Set the position coordinates of the previous optimal viewpoint to , the position coordinates of the next optimal viewpoint are ; Set the drone speed to , the optimal track angle is And the optimal heading angle is , the trajectory planning between two adjacent optimal viewpoints is expressed as:

[0027] ;

[0028] ;

[0029] in, represents the objective function of optimal trajectory planning, 、 、 are the velocity components of the UAV in the x, y, and z coordinate axes, respectively. The integration interval is from t = 0 to t = k, indicating that the time it takes for the UAV to fly from the previous optimal viewpoint to the next adjacent optimal viewpoint is k; 、 、 They are the offsets in the x, y, and z coordinate axes, respectively. Here, it is the wind speed attached to the drone. () is the four-quadrant inverse tangent function; 、 、 Indicates the position coordinates of the previous optimal viewpoint on the x, y, and z axes at time t = 0 、 、 Indicates the position coordinates of the next optimal viewpoint on the x, y, and z axes at time t=k. , then the optimal heading angle , construct the system Hamiltonian function through real-time wind speed measurement by UAV:

[0030] ;

[0031] in, 、 、 They are Lagrange multipliers respectively, and are calculated by the Hamiltonian function H of the system. and Calculate the partial derivatives respectively and set them equal to 0 to find the optimal track angle and optimal heading angle ;

[0032] The optimal track path between two adjacent optimal viewpoints is obtained according to the optimal track angle and the optimal heading angle;

[0033] The optimal trajectory paths are acquired for adjacent optimal viewpoints in the global inspection path in sequence to obtain the optimized global inspection path.

[0034] Furthermore, the position to be inspected of the climbing robot in the minimum convex polygon is determined according to the global inspection result, specifically:

[0035] Obtain the inspection location set of the drone based on the global inspection results;

[0036] Obtaining the climbing angle and the descending angle of the climbing robot;

[0037] The positions to be inspected are searched in the inspection position set according to the climbing angle and the descending angle, the positions to be inspected of the climbing robot are obtained, and a set of positions to be inspected is constructed.

[0038] Furthermore, an improved branch and bound algorithm is used to determine the vertical inspection path of the climbing robot, including:

[0039] Generate an initial inspection path for the climbing robot using a greedy algorithm according to the set of positions to be inspected;

[0040] The initial feasible solution of the greedy algorithm is used as the initial upper bound;

[0041] Starting from the first position to be inspected in the set of positions to be inspected, the task points are expanded between the next positions to be inspected;

[0042] Calculate the search tree upper bound of the task point. If the search tree upper bound is greater than the initial upper bound, re-expand the task point. If the search tree upper bound is less than the initial upper bound, use the task point as the next position to be inspected.

[0043] The task points are expanded again starting from the next position to be inspected, and the vertical inspection path is obtained by iteration.

[0044] The beneficial effects of the present invention are: the inspection planning provided by the present invention includes two parts: UAV visual inspection planning and climbing robot contact inspection planning; among them, the UAV visual inspection planning constructs a three-dimensional point cloud description of the vertical structure facility through a three-dimensional convex hull incremental algorithm, obtains the inspection viewpoint of the vertical structure facility, and completes the UAV viewpoint inspection based on the viewpoint description using the improved A-star algorithm, thereby meeting the minimum inspection path planning requirements; the climbing robot contact inspection planning is based on the inspection results of the UAV, thereby constructing a rolling branch and bound algorithm to complete the inspection, and through this algorithm, meets the increasing inspection tasks based on the UAV, performs real-time inspection planning adjustments, and meets the inspection quality while also meeting the requirements of minimizing the inspection path. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 This is a flow chart of a method for planning a coordinated three-dimensional inspection of heterogeneous robots in a vertical structure facility according to an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of a viewpoint acquisition principle according to an embodiment of the present invention;

[0048] Figure 3 This is a flow chart of a method for planning a UAV inspection path according to an embodiment of the present invention;

[0049] Figure 4 The figure is a flow chart of a method for planning an inspection path for a climbing robot according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] A flowchart of a method for planning a coordinated three-dimensional inspection of heterogeneous robots in a vertical structure facility according to an embodiment of the present invention is shown in FIG. Figure 1 As shown, including:

[0052] Acquire point cloud data of a target facility with a vertical structure; and generate a minimum convex polygon describing the target facility with a vertical structure using a three-dimensional convex hull enhancement algorithm based on the point cloud data;

[0053] In the embodiment of the present invention, in order to realize the inspection of the target facility with vertical structure, a point cloud is first generated on the target facility with vertical structure: given the projection information and height information of the target facility with vertical structure, the circumference angle of the target facility with vertical structure is evenly divided into 2n equal parts to obtain the azimuth angle. ,n= , with the centroid as the starting point and the azimuth as Draw rays and generate a point at the boundary of the target facility of the vertical structure on each divided ray. If the height is given, 2n point clouds are generated at the given height. The height of the target facility of the vertical structure is divided into m equal parts of the given height, and the number of point clouds generated is m×2n.

[0054] The embodiment of the present invention further adopts a three-dimensional convex hull enhancement algorithm to achieve the minimum convex polygon description of the vertical structure target facility. The three-dimensional convex hull enhancement algorithm can be expressed as:

[0055] ;

[0056] in, To solve the minimum volume convex polygon, and They represent the volume of the selected convex polygon and the target facility volume of the vertical structure respectively. The specific implementation steps are as follows:

[0057] Select three non-collinear points p1, p2 and p3 in the point cloud as the three vertices of the initial convex hull; traverse the point cloud: for any point p in the point cloud, if point p is inside the initial convex hull, skip it, otherwise traverse all edges of the initial convex hull to determine whether point p is outside some faces. If so, delete the face; for all edges of the initial convex hull, check whether point p is on one side of the edge. If so, delete the face and add two new faces with point p and the endpoints on the edge as vertices; add point p to the initial convex hull: when all points in the point cloud are traversed, the three-dimensional convex hull is obtained, and a point cloud set of the describable three-dimensional convex hull is obtained; based on the point cloud set of the three-dimensional convex hull, the aerial drone constructs corresponding viewpoints to create a coverage path, and the climbing robot constructs a random traveling salesman problem based on the real-time changing inspection requirements to complete the inspection task.

[0058] Determine a cruising start point and a cruising end point of the UAV based on the minimum convex polygon;

[0059] The embodiment of the present invention obtains the inspection viewpoint of the vertical structure target facility based on the minimum convex polygon description of the vertical structure target facility. Specifically, in the horizontally divided sub-area, based on the task division result, the safe distance between the drone and the facility wall is given, and the regional grid is used to obtain the inspection viewpoint of the vertical structure target facility. , get the corresponding viewpoint , expressed as:

[0060] ;

[0061] in, represents a grid, For Grid The normal vector of , i represents the i-th grid cell, , which represents a method for determining Relative in direction The offset, d is the required sampling distance, which represents the safe distance between the drone and the wall of the vertical structure target facility. Detect width for the visual sensor installed on the drone, is the horizontal field of view of the visual sensor installed on the drone, and a viewpoint acquisition method of an embodiment of the present invention is as follows: Figure 2 As shown, based on the viewpoint acquisition method, the cruising start point and cruising end point of the UAV on the vertical structure target facility can be further determined according to the actual mission requirements.

[0062] The improved A-star algorithm is used to plan the path between the UAV's cruise start point and cruise end point to determine the UAV's global inspection path;

[0063] In the embodiment of the present invention, the drone inspection adopts the improved A-star algorithm (i.e., A* algorithm). The algorithm plans the path by introducing optimization measures such as collision prediction, adaptive step size adjustment, and collision threat cost function, and safely and effectively passes through the vertical structure facility environment to complete the inspection coverage. The flow diagram is shown below. Figure 3 As shown, by setting a heuristic function, the heuristic function is used to estimate the cost of the UAV from the cruise starting point to the cruise end point; constructing a cost function according to the turning cost, rising cost and collision avoidance cost of the UAV from the cruise starting point to the cruise end point; starting from the cruise starting point, searching for the next viewpoint, calculating the priority of the next viewpoint according to the heuristic function and the cost function, and selecting the next viewpoint with the highest priority as the optimal viewpoint; taking the optimal viewpoint as the starting point, searching for the next viewpoint again, and selecting the next viewpoint with the highest priority as the optimal viewpoint; iterating in sequence until reaching the cruise end point, and obtaining all the optimal viewpoints; obtaining the global inspection path of the UAV according to all the optimal viewpoints.

[0064] The details are as follows:

[0065] (1) Setting a heuristic function: The heuristic function h(n) is used to estimate the cost from each viewpoint to the next viewpoint starting from the cruise starting point. In a vertical structure, the Euclidean distance can be used as the heuristic function. The embodiment of the present invention provides a method for calculating the cost from the current viewpoint to the next optimal viewpoint, namely:

[0066] ;

[0067] in, is the current viewpoint, is the next optimal viewpoint, ( ) is the coordinate of the current viewpoint on the x, y, and z axes, ( ) is the coordinate of the next optimal viewpoint on the x, y, and z axes.

[0068] (2) Setting the cost function: The cost function g(n) represents the actual cost from each viewpoint to the next viewpoint starting from the cruise starting point. In the vertical structure, the embodiment of the present invention constructs the cost function by considering the UAV's turning cost, ascent cost, and collision avoidance cost. Taking the actual cost from the current viewpoint to the next viewpoint as an example, its cost function can be expressed as:

[0069] ;

[0070] Where d represents the Euclidean distance from the current viewpoint to the next optimal viewpoint, , , are the cost coefficients, ; Turn cost It is calculated based on the angle between the current path and the previous path, that is:

[0071] ;

[0072] in, and Represent the direction vectors of the current segment and the next segment respectively, where the current segment refers to any next viewpoint with the current viewpoint as the previous optimal viewpoint, that is, the current viewpoint may also be the next optimal viewpoint of the previous optimal viewpoint, then the path from the previous optimal viewpoint to the current viewpoint is the current segment, and the path between the current segment and any next viewpoint is the next segment; the rising cost Calculated based on height difference, that is: Safety cost To avoid approaching obstacles, the distance between the current viewpoint and the obstacle can be calculated as the inverse of the distance, i.e. ;in, Indicates the position coordinates of the current viewpoint, Indicates the location coordinates of the obstacle. The closer the distance between the current viewpoint and the obstacle, the higher the safety cost.

[0073] Starting from the cruise starting point, the improved A-star algorithm is used to search for paths. In each iteration, the comprehensive priority f(n)=g(n)+h(n) of each viewpoint is calculated according to the heuristic function and the cost function, and the viewpoint with the highest priority is selected for expansion. After determining the optimal viewpoint for expansion, it is necessary to plan the route between the two points. For three-dimensional inspection, UAV inspection faces the problem of wind shear caused by the interaction between the wind field and the vertical structure, which in turn causes turbulence. Therefore, the embodiment of the present invention further provides a UAV dynamics description including a wind field description, and then performs online correction on the inspection trajectory. In the track planning between two given viewpoints, the position coordinates of the previous optimal viewpoint are , the position coordinates of the next optimal viewpoint are , define the drone speed as , the optimal track angle is And the optimal heading angle is , then the optimal trajectory planning is expressed as:

[0074] ;

[0075] ;

[0076] in, represents the objective function of optimal trajectory planning, 、 、 are the velocity components of the UAV in the x, y, and z coordinate axes, respectively. The integration interval is from t = 0 to t = k, indicating that the time it takes for the UAV to fly from the previous optimal viewpoint to the next adjacent optimal viewpoint is k; 、 、 They are the offsets in the x, y, and z coordinate axes, respectively. Here, it is the wind speed attached to the drone. () is the four-quadrant inverse tangent function; 、 、 Indicates the position coordinates of the previous optimal viewpoint on the x, y, and z axes at time t = 0 、 、 Indicates the position coordinates of the next optimal viewpoint on the x, y, and z axes at time t=k. , then the optimal heading angle , construct the system Hamiltonian function through real-time wind speed measurement by UAV:

[0077] ;

[0078] in, 、 、 They are Lagrange multipliers respectively, and are calculated by the Hamiltonian function H of the system. and Calculate the partial derivatives respectively and set them equal to 0 to find the optimal track angle and optimal heading angle , the path correction can be achieved, so that the UAV can reach the next optimal viewpoint position from the initial viewpoint position in the shortest time, and the inspection path trajectory planning can be achieved, that is, the global inspection path of the embodiment of the present invention is obtained.

[0079] The drone conducts inspections according to the determined global inspection path and obtains global inspection results;

[0080] Drone inspections are easily affected by external environmental factors such as wind field interference, which may lead to missed inspections. In addition, for specific areas in vertical facilities where scene information is insufficient, such as valves, pipeline joints and suspected ruptures in storage tanks, climbing robots are needed to conduct detailed inspections one by one to achieve information-enhanced perception.

[0081] Therefore, the embodiment of the present invention uses the drone online inspection to obtain the local surface boundary, and the climbing robot extracts the surface boundary from the shared map information. Secondly, as the drone online inspection progresses, through information sharing, the robot will gradually add the missed inspection area, forming an inspection planning problem under uncertainty, thereby obtaining the global inspection result; at this time, the inspection task set consists of a set of known inspection locations. and a set of uncertain inspection locations There are m fixed positions, represented by a set S, which are 、 Until , the other type is uncertain inspection location, using the set Indicates that it contains nm positions, starting from arrive .

[0082] Determining a position to be inspected of the climbing robot in the minimum convex polygon according to the global inspection result;

[0083] A heterogeneous robot system usually integrates multiple different types of robots, and works together to complete complex tasks that are difficult for a single type of robot to complete. In the embodiment of the present invention, the heterogeneous robot is a structural collection of drones and climbing robots. The drones and climbing robots perform collaborative inspections to achieve precise inspections under vertical structures. Therefore, after determining the global inspection path of the drone, the inspection path of the climbing robot must also be determined.

[0084] The embodiment of the present invention sets the initial position of the climbing robot to , in the inspection position To the inspection location The distance is , climbing angle of the climbing robot and descent angle Constraints, i.e. ,Right now In order to climb the minimum and maximum deviation angles, the problem is constructed as a random traveling salesman variant with a fixed starting point. Considering that each inspection location is visited and each location is inspected only once, the covering inspection path of the climbing robot based on the global inspection path can be established as follows:

[0085] ;

[0086] st ;

[0087] in, Indicates the total length of the inspection path covered. Indicates that the inspection position Go to the next inspection location distance, is the initial position of the climbing robot and also the final position. Is a permutation, indicating the order of visiting inspection locations, Indicates inspection location To the end position of the climbing robot distance, Indicates the initial position of the climbing robot Go to the next unvisited inspection location to be visited The distance from the initial position of the climbing robot Start by getting the next unvisited inspection location to be visited The distance to the next unvisited inspection location to be visited is obtained in sequence , until the last unvisited inspection location to be visited , and finally return to the final position of the climbing robot ; n is the total number of inspection locations; minimizing the maximum path length of the inspection path planning is taken as the main goal of the coverage inspection. At this time, the coverage inspection path includes many locations that are not inspected in the global inspection path, which are the paths to be inspected by the climbing robot.

[0088] According to the location to be inspected, an improved branch and bound algorithm is used to determine the vertical inspection path of the climbing robot.

[0089] The embodiment of the present invention adopts an improved branch and bound algorithm to optimize the climbing transition gait to minimize the inspection path. The flow chart is as follows: Figure 4 As shown, it is implemented according to the following steps:

[0090] (1) Initialization: Given a set of known inspection locations }, the drone missed the inspection location during the initial process , the starting position of the climbing robot , given the minimum climbing offset angle and the maximum offset angle.

[0091] (2) Taking the known inspection locations as the inspection target, the greedy algorithm is used to generate the initial path. Based on the initial path, the nearest inspection location is gradually selected until all known inspection locations are inspected. At this time, an initial feasible solution is generated as the current upper bound. , the lower bound It is obtained by using the linear programming relaxation algorithm, namely:

[0092] ;

[0093] ;

[0094] Among them, here represents the distance from inspection position i to inspection position j, Is a binary variable. When moving directly from inspection position i to inspection position j, =1; otherwise =0, the goal is to minimize the total distance of all inspection paths; use the simplex method to solve the relaxation problem, is the total number of inspection locations, and the constraint Indicates that for each inspection location j, there is only one path to enter the inspection location, and the constraint It means that for each inspection location i, there is only one path starting from the inspection location, that is, the first and second constraints mean that each inspection location is entered exactly once and each inspection location is left exactly once. The third constraint means eliminating the sub-loop constraint, which means that for i and j, the value range is , and i is not equal to j, the inequality is satisfied when When , the inequality becomes ;when When , the inequality becomes ,Right now In the traveling salesman problem (TSP, the inspection path planning problem here can be regarded as a variant of TSP), there may be some small loops that do not contain all inspection locations. This constraint can avoid this situation and ensure that a complete path traversing all inspection locations is obtained.

[0095] (3) Inspection task starts from the initial position of the climbing robot Start to gradually expand the task points, starting from the last inspection position of the current path, considering the angle relationship, and select an unvisited inspection position as the next unvisited inspection position to be visited , calculate the height difference between the two inspection locations , and calculate the horizontal distance between the two inspection positions:

[0096] ;

[0097] Where, Indicates the horizontal distance between two inspection positions. Indicates that it is selected as the next unvisited inspection location to be visited. Indicates the difference between the last inspection position s on the current path and the next unvisited inspection position to be visited The horizontal distance between 、 Indicates the x-axis and y-axis coordinates of the last inspection position of the current path in the plane rectangular coordinate system. 、 Indicates the next unvisited inspection location to be visited The x-axis and y-axis coordinates in a rectangular coordinate system.

[0098] Calculate the offset angle , check whether the angle meets the constraint conditions. If the path does not meet the angle constraint, it means that the task point is currently unreachable; if the constraint conditions are met, based on the distance between the two inspection locations in the current path ,by Calculate the upper bound of the search tree for the next task point ,like , then perform pruning and ignore the newly constructed local feasible solution; if , then update the upper bound for ; Use the relaxation algorithm to calculate the lower bound ,like , then perform pruning operations to remove all local feasible solutions; iteratively search all task points in turn, and when the search tree traversal is completed, return the current upper bound The corresponding path is used as the optimal path, and the smooth path planning is completed by using the cubic Bezier curve. The embodiment of the present invention determines the intermediate control points according to the surface shape and the robot motion constraints. , then the Bezier curve is described as:

[0099] ;

[0100] in, 、 Indicates the starting and ending points of the path, and the intermediate control points Should be in the direction of the starting point tangent, the middle control point It should be in the opposite direction of the end point tangent.

[0101] ;

[0102] ;

[0103] Among them, k is a scaling factor that controls the curvature of the curve. is the unit direction vector of the tangent line at the starting point, is the unit direction vector of the tangent line at the end point.

[0104] (4) During the inspection process, the drone may find new missed inspection areas and generate new uncertain tasks. For any new task points, they are added to the task set. The added task only affects the paths that have not been inspected, and has no effect on the paths that have been inspected. Therefore, a greedy algorithm is used to quickly generate new feasible solutions and update the upper bound. Recalculate the lower bound using the linear programming relaxation algorithm ,The inspection process continues and the climbing UAV starts from the current task point, ,expands the task point, still considering the angle constraint for pruning ,operations, and performs further pruning operations based on the upper and lower bound ,calculations, and finally returns the updated path.

[0105] The improved branch and bound method in the embodiment of the present invention can effectively solve the path planning problem of drones and climbing robots during the inspection process through dynamic upper and lower bound updates, incremental search and angle constraint integration. While ensuring accuracy, this method has strong dynamic adjustment capabilities and computational efficiency, and is suitable for practical application scenarios.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for planning a coordinated three-dimensional inspection of heterogeneous robots in vertically structured facilities, characterized in that: include: Acquire point cloud data of target facilities with vertical structures; A three-dimensional convex hull enhancement algorithm is used based on the point cloud data to generate the minimum convex polygon of the target facility describing the vertical structure; specifically: Select three non-collinear point clouds in the target facility point cloud data of the vertical structure as three vertices to establish the initial convex hull; Select any point cloud in the target facility point cloud data of the vertical structure and determine whether the point cloud is inside the initial convex hull. If it is inside, skip it. If it is not inside, determine the positional relationship between the point cloud and each edge of the initial convex hull, and add the point cloud to the initial convex hull based on the positional relationship; All point clouds in the target facility point cloud data of the vertical structure are traversed in sequence to obtain a convex hull point cloud set of the target facility that describes the vertical structure; According to the convex hull point cloud set, the minimum convex polygon of the target facility describing the vertical structure is obtained; Determine a cruising start point and a cruising end point of the UAV based on the minimum convex polygon; The improved A-star algorithm is used to plan the path between the UAV's cruise start point and cruise end point to determine the UAV's global inspection path; The drone performs inspection according to the determined global inspection path and obtains a global inspection result; Determining a position to be inspected of the climbing robot in the minimum convex polygon according to the global inspection result; According to the position to be inspected, an improved branch and bound algorithm is adopted to determine the vertical inspection path of the climbing robot.

2. The method for planning a coordinated three-dimensional inspection of a vertically structured facility by heterogeneous robots according to claim 1, characterized in that: The improved A-star algorithm is used to plan the path between the UAV's cruise start point and cruise end point, including: Setting a heuristic function, wherein the heuristic function is used to estimate the cost of the UAV from the cruise start point to the cruise end point; Construct a cost function based on the turning cost, rising cost and collision avoidance cost of the UAV from the cruising start point to the cruising end point; Searching for a next viewpoint from the cruise starting point, calculating the priority of the next viewpoint according to the heuristic function and the cost function, and selecting the next viewpoint with the highest priority as the optimal viewpoint; Taking the optimal viewpoint as the starting point, search for the next viewpoint again and select the next viewpoint with the highest priority as the optimal viewpoint; Iterate in sequence until the cruise end is reached and all optimal viewpoints are obtained; The global inspection path of the UAV is obtained based on all the optimal viewpoints.

3. The method for planning a coordinated three-dimensional inspection of a vertically structured facility by heterogeneous robots according to claim 2, characterized in that: When obtaining the global inspection path of the UAV based on all the optimal viewpoints, it also includes: Set the position coordinates of the previous optimal viewpoint to , the position coordinates of the next optimal viewpoint are ; Set the drone speed to , the optimal track angle is And the optimal heading angle is , the trajectory planning between two adjacent optimal viewpoints is expressed as: ； ； in, represents the objective function of optimal trajectory planning, 、 、 are the velocity components of the UAV in the x, y, and z coordinate axes, respectively. The integration interval is from t = 0 to t = k, indicating that the time it takes for the UAV to fly from the previous optimal viewpoint to the next adjacent optimal viewpoint is k; 、 、 They are the offsets of the wind speed attached to the drone in the directions of the x, y, and z coordinate axes, is the four-quadrant inverse tangent function; 、 、 Indicates the position coordinates of the previous optimal viewpoint on the x, y, and z axes at time t = 0, 、 Represents the position coordinates of the next optimal viewpoint on the x, y, and z axes at time t=k. If , then the optimal heading angle The value increases by one cycle ; Construct the system Hamiltonian function through real-time wind speed measurement by drone: ； in, 、 、 They are Lagrange multipliers respectively, and are calculated by the Hamiltonian function H of the system. and Calculate the partial derivatives respectively and set them equal to 0 to find the optimal track angle and optimal heading angle ; The optimal track path between two adjacent optimal viewpoints is obtained according to the optimal track angle and the optimal heading angle; The optimal trajectory paths are acquired for adjacent optimal viewpoints in the global inspection path in sequence to obtain the optimized global inspection path.

4. The method for planning a coordinated three-dimensional inspection of a vertically structured facility by heterogeneous robots according to claim 1, wherein: The position to be inspected of the climbing robot in the minimum convex polygon is determined according to the global inspection result, specifically: Obtain the inspection location set of the drone based on the global inspection results; Obtaining the climbing angle and the descending angle of the climbing robot; The positions to be inspected are searched in the inspection position set according to the climbing angle and the descending angle, the positions to be inspected of the climbing robot are obtained, and a set of positions to be inspected is constructed.

5. The method for planning a coordinated three-dimensional inspection of a vertically structured facility by heterogeneous robots according to claim 4, characterized in that: An improved branch-and-bound algorithm is used to determine the vertical inspection path of the climbing robot, including: Generate an initial inspection path for the climbing robot using a greedy algorithm according to the set of positions to be inspected; The initial feasible solution of the greedy algorithm is used as the initial upper bound; Starting from the first position to be inspected in the set of positions to be inspected, the task points are expanded between the next positions to be inspected; Calculate the search tree upper bound of the task point, if the search tree upper bound is greater than the initial upper bound, re-expand the task point; if the search tree upper bound is less than the initial upper bound, use the task point as the next position to be inspected; The task points are expanded again starting from the next position to be inspected, and the vertical inspection path is obtained by iteration.

Images