Substation unmanned aerial vehicle autonomous inspection route generation method based on optimization A star algorithm

Abstract

The invention belongs to the technical field of unmanned aerial vehicles, and particularly provides a transformer substation unmanned aerial vehicle autonomous inspection route generation method based on an optimization A star algorithm. According to the method, a heuristic function is improved, the Euclidean distance between a target point and a current node is comprehensively considered, obstacle avoidance processing is carried out, and a safer and more efficient inspection route is generated. And by calculating the total cost of each node, the node with the minimum total cost is selected for expansion until a target node is found. According to the improved A star algorithm, the unmanned aerial vehicle can be effectively prevented from flying into an obstacle dense area, the obstacle keeps a safe distance, and the flying safety is remarkably improved. The distance and the obstacle density are comprehensively considered, the generated inspection route is smoother and more efficient, and unnecessary flight distance and energy consumption are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle inspection, and in particular to a method for generating a substation unmanned aerial vehicle autonomous inspection route based on an improved A-star algorithm. Background Art

[0002] With the development of drone technology, drones have been widely used in many fields. However, in the complex substation environment, the path planning and obstacle avoidance of drones have become one of the key issues restricting their widespread application. Current path planning algorithms such as the A algorithm have been widely used in two-dimensional plane path planning, but directly applying the traditional A algorithm in three-dimensional space faces the problems of low computational efficiency, non-optimized path, and inability to avoid obstacles in real time.

[0003] In the three-dimensional space obstacle avoidance and path planning of UAVs, the current technical solutions mainly include the following categories:

[0004] 1. Preliminary application of three-dimensional path planning algorithms. For the application of drones in three-dimensional space, some studies have extended two-dimensional algorithms to three dimensions, such as the three-dimensional A algorithm. This algorithm enables path planning in three-dimensional space by adding a dimension. However, the traditional three-dimensional A algorithm has low computational efficiency in complex environments, especially when obstacles are dense or large in area. The amount of calculation increases exponentially, which easily leads to a long path planning time and is not suitable for tasks with high real-time requirements.

[0005] 2. Obstacle avoidance method based on potential field. The potential field method sets the target point as the attraction source and the obstacle as the repulsion source to guide the drone to move along the gradient direction of the potential field. However, the potential field method has the problem of local minima, that is, the drone may fall into some areas where it cannot reach the target. In addition, when dealing with complex obstacle environments, the potential field method often finds it difficult to provide an optimized path and is prone to oscillation in areas with dense obstacles.

[0006] 3. Path planning based on RRT (Rapidly Exploring Random Trees). The RRT algorithm is an algorithm widely used in robot path planning, which can quickly generate a path from the starting point to the target point. The advantage of the RRT algorithm is that it can handle path planning problems in high-dimensional space, and is particularly suitable for complex three-dimensional environments. However, the paths generated by RRT are usually more random, with problems of path redundancy and unevenness, and the optimality of the path cannot be guaranteed. During the obstacle avoidance process, RRT also finds it difficult to respond to dynamic changes in the environment in real time.

[0007] 4. Real-time dynamic obstacle avoidance. The real-time dynamic obstacle avoidance technology of drones usually relies on the real-time perception of the environment by sensors (such as LiDAR, cameras), and dynamically adjusts the flight path through local path planning algorithms (such as DWA). This type of technology can effectively deal with sudden obstacles, but in complex environments, local path planning may fall into a local minimum, causing the drone to be unable to move forward or fall into a repetitive loop. In addition, dynamic obstacle avoidance is usually separated from global path planning, which is prone to path discontinuity or multiple adjustments, affecting the flight stability of the drone.

[0008] The existing UAV inspection path planning technology has the following shortcomings:

[0009] 1. Low computational efficiency: Traditional 3D path planning algorithms (such as the 3D A-star algorithm) have huge computational complexity when dealing with complex 3D environments and cannot meet real-time requirements.

[0010] 2. Unsmooth paths: The paths generated by algorithms like RRT in three-dimensional space are often not smooth enough, with many redundant nodes and sharp turns, which affect the flight stability and efficiency of the drone.

[0011] 3. Local minimum problem: The potential field method and the local dynamic obstacle avoidance method are prone to fall into local minima in complex environments, resulting in path planning failure or inefficiency.

[0012] 4. Lack of global and local path planning: Existing real-time dynamic obstacle avoidance technology is often separated from global path planning, resulting in poor path continuity and overall optimization, affecting the flight performance of drones in complex environments. Summary of the invention

[0013] The purpose of the present invention is to provide a method for generating autonomous inspection routes for substation drones based on an optimized A-star algorithm to solve the problems existing in the prior art. The present invention improves the A-star algorithm and combines the characteristics of the substation environment to achieve efficient and real-time path planning and obstacle avoidance.

[0014] The technical solution of the present invention is as follows:

[0015] A method for generating autonomous inspection routes for substation drones based on an optimized A-star algorithm is characterized in that the nodes in the inspection route planning path are generated by adopting the following steps: calculating the evaluation function f(n) of the node, and in each step of expanding the route node, always selecting the node with the smallest function value of the evaluation function f(n) as the child node, and then using the child node as the parent node to expand and optimize the next step of the route child node, that is, sequentially searching for the child node with the smallest value of the evaluation function f(n) to form an optimized route;

[0016] The expression of the evaluation function f(n) is:

[0017] f(n)=g(n)+α[h(n)+h(p)]

[0018] Where: g(n) is the actual route cost from the starting node to the current node; h(n) is the estimated route cost from the current node to the target point; h(p) is the distance from the parent node of the current child node to the target point; α is the weight value.

[0019] The method for generating a substation unmanned aerial vehicle autonomous inspection route based on an optimized A-star algorithm is characterized in that the starting point is placed in a CLOSE list as the first node, and the evaluation functions of up to 8 surrounding nodes are calculated. Due to flight boundaries and obstacles, the number of surrounding nodes may be less than 8, and these nodes and node information are placed in an OPEN list. If the current node already exists in the OPEN list, the evaluation function value of the node is compared, and the node information with a smaller evaluation function value is updated to the OPEN list; in the OPEN list, the node with the smallest evaluation function value is selected as the child node of the current node; then the selected child node is used as the parent node, and its parameter information is placed in the CLOSE list, and the new parent node is used to expand the child node and calculate the evaluation function value of each child node, and a new child node is obtained by comparing the evaluation function value, and the CLOSE and OPEN lists are continuously updated until the target point is searched.

[0020] The method for generating autonomous inspection routes for substation drones based on the optimized A-star algorithm is characterized in that when expanding child nodes, if there are obstacles in the spatial direction of the current node, the nodes on the grid diagonal are not expanded as child nodes, that is, diagonal routes with a distance from the obstacle less than 1 grid unit are not generated.

[0021] The method for generating a substation drone autonomous inspection route based on the optimized A-star algorithm is characterized in that g(n) is the actual route cost from the starting node to the current node; h(n) is the estimated route cost from the current node to the target point, and the Euclidean distance is used as the calculation method of the route cost, and its expression is:

[0022]

[0023]

[0024] Where: (x i ,y i ) is the current point coordinate; (x s ,y s ) is the starting point coordinate; (x t ,y t ) are the target point coordinates.

[0025] The method for generating a substation drone autonomous inspection route based on an optimized A-star algorithm is characterized in that, taking a two-dimensional space environment as an example, when expanding child nodes, if there are obstacles in the X positive, X negative, Y positive, and Y negative directions of the current node, the nodes on the grid diagonal are not expanded as child nodes, that is, a diagonal route with a distance from the obstacle less than 1 grid unit is not generated.

[0026] The method for generating a substation drone autonomous inspection route based on the optimized A-star algorithm is characterized in that the generated optimal route is simplified by combining the A-star algorithm and the Floyd algorithm, and multiple iterative optimizations are performed from the starting point to the target point using the Floyd algorithm, thereby shortening the length of the planned route.

[0027] The method for generating a substation drone autonomous inspection route based on the optimized A-star algorithm is characterized in that the obstacle density is taken into consideration. If the obstacle density near the current node is high, the heuristic cost is increased to prevent the drone from flying into an obstacle-dense area.

[0028] h′(n)=h(n)+w×density(n);

[0029] Among them, w is the weight coefficient, and density(n) represents the obstacle density near node n.

[0030] The method for generating a substation drone autonomous inspection route based on the optimized A-star algorithm is characterized in that when the planned path contains unnecessary turning points and sharply changing directions, the following processing is performed:

[0031] Use B-Spline to smooth the planned path to make it smoother and more natural;

[0032]

[0033] Where C(t) is the coordinate of the curve, Pi is the control point, Ni,p(t) is the B-spline basis function of the ith control point, and its order is p;

[0034] The main characteristics of B-splines include:

[0035] The curve is controlled by the convex hull of the control points;

[0036] The shape and smoothness of the curve are determined by the locations of the control points;

[0037] The curve is a polynomial in each segment, and local adjustments can be made through control points without affecting the overall curve.

[0038] The advantages of the present invention are as follows:

[0039] Based on the traditional A-star algorithm, an improved heuristic function is designed according to the characteristics of substations. This algorithm combines factors such as distance in space, keeping a safe distance from obstacles, and obstacle density to improve the efficiency and safety of path planning.

[0040] Path search and optimization: Use the improved A-star algorithm for initial path search, and perform smoothing and path optimization on the generated initial path to reduce the turning points and total flight distance of the path.

[0041] Improved safety: By considering obstacle density, the improved A-star algorithm can maintain a safe distance from obstacles and prevent the drone from flying into areas with dense obstacles, thereby reducing the risk of collision and improving flight safety.

[0042] Improved efficiency: Although the introduction of obstacles to maintain a safe distance and the obstacle density term increases the computational complexity of the heuristic function, this improvement enables the algorithm to more quickly eliminate paths with high obstacle density during the search process, thereby reducing unnecessary searches and calculations and improving the efficiency of path planning.

[0043] Improved path quality: By taking into account obstacle density, the improved A-star algorithm can find a smoother and safer path, which is beneficial for the flight control and energy consumption of the drone.

[0044] In three-dimensional space path planning, the present invention takes into account the influence of keeping a safe distance from obstacles and obstacle density on the flight of drones, and can improve the traditional A-star algorithm heuristic function. The improved heuristic function not only considers the Euclidean distance between the target point and the current point, but also introduces items related to keeping a safe distance from obstacles and obstacle density. The algorithm tends to choose paths with lower obstacle density during the search process. This improvement enables the algorithm to more comprehensively evaluate the costs of different paths during the search process, thereby preventing drones from flying into areas with dense obstacles. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is the flow chart of the A-star algorithm optimized for the present invention.

[0046] Figure 2 Improve the route map of the former A-star algorithm for the evaluation function.

[0047] Figure 3 This is the route diagram of the A-star algorithm after the evaluation function is improved.

[0048] Figure 4 To improve the expansion of child node mode diagram.

[0049] Figure 5 Improved route generation for extended child nodes.

[0050] Figure 6 This is Floyd's shortest route principle. DETAILED DESCRIPTION

[0051] Example 1:

[0052] The following is a detailed description of each step of the substation drone autonomous inspection route generation method based on the optimized A-star algorithm, including relevant algorithms, formulas, and implementation logic descriptions.

[0053] A method for generating autonomous inspection routes for substation drones based on optimized A-star algorithm.

[0054] The nodes in the inspection route planning path are generated by the following steps: the evaluation function f(n) of the node is calculated. In each step of expanding the route node, the node with the smallest function value of the evaluation function f(n) is always selected as the child node, and then the child node is used as the parent node to expand and optimize the next route child node, that is, the child node with the smallest value of the evaluation function f(n) is searched in turn to form an optimal route;

[0055] The expression of the evaluation function f(n) is:

[0056] f(n)=g(n)+α[h(n)+h(p)]

[0057] Where: g(n) is the actual route cost from the starting node to the current node; h(n) is the estimated route cost from the current node to the target point; h(n) is the distance from the parent node of the current child node to the target point; α is the weight value.

[0058] g(n) is the actual route cost from the starting node to the current node; h(n) is the estimated route cost from the current node to the target point. Taking two-dimensional space as an example, the Euclidean distance is used as the calculation method of the route cost, and its expression is:

[0059]

[0060]

[0061] Where: (x i ,y i ) is the current point coordinate; (x s ,y s ) is the starting point coordinate; (x t ,y t ) are the target point coordinates.

[0062] Examples of the A-star algorithm before and after improving the evaluation function Figure 2 and Figure 3 As shown in Table 1, compared with the conventional A-star algorithm, its route length and number of search nodes are compared.

[0063] from Figure 3 It can be seen that after the evaluation function is improved, the number of search nodes is reduced by nearly 25%, and the number of closed-loop nodes is reduced by nearly 55%. That is, the number of useless round trips is significantly reduced, and the search efficiency of the algorithm is improved. At the same time, the range of weight variation is not large. When it exceeds a certain range, the change in weight has little effect on the route search results.

[0064] Table 1 Comparison of the number of nodes and route length of different methods

[0065]

[0066] Table 1 Comparison of the number of nodes and route length of different methods

[0067] Put the starting point as the first node into the CLOSE list, calculate the evaluation function of up to 8 surrounding nodes. Due to flight boundaries and obstacles, the number of surrounding nodes may be less than 8, and put these nodes and node information into the OPEN list. If the current node already exists in the OPEN list, compare the evaluation function value of the node, and update the node information with a smaller evaluation function value to the OPEN list; in the OPEN list, select the node with the smallest evaluation function value as the child node of the current node; then take the selected child node as the parent node, and put its parameter information into the CLOSE list, use the new parent node to expand the child nodes and calculate the evaluation function value of each child node, obtain new child nodes by comparing the evaluation function values, and continuously update the CLOSE and OPEN lists until the target point is found.

[0068] Example 2

[0069] A method for generating autonomous inspection routes for substation drones based on the optimized A-star algorithm. When expanding child nodes, if there are obstacles in the spatial direction of the current node, the nodes on the grid diagonal will not be expanded as child nodes, that is, the diagonal routes with a distance less than 1 grid unit from the obstacles will not be generated.

[0070] Taking a two-dimensional space environment as an example, when expanding child nodes, if there are obstacles in the positive X, negative X, positive Y, and negative Y directions of the current node, the nodes on the grid diagonal will not be expanded as child nodes, that is, diagonal routes that are less than 1 grid unit away from obstacles will not be generated.

[0071] like Figure 2 As shown in the figure, assuming that the current waypoint is A→B, there is an obstacle node on the upper side of waypoint B. According to the conventional A-star extended subnode algorithm, the best extended subnode is B→C→D, but the route is less than 1 grid unit away from the obstacle node (4, 2). Therefore, when B expands the subnode, it is not allowed to generate the subnode C. The best subnodes should be B→E→C→D in sequence.

[0072] That is, when there is an obstacle node in the positive X direction of the current node, it is not allowed to generate child nodes of X positive Y positive or X positive Y negative; when there is an obstacle node in the negative X direction, it is not allowed to generate child nodes of X negative Y positive or X negative Y negative; when there is an obstacle node in the positive Y direction, it is not allowed to generate child nodes of Y positive X positive or Y positive X negative; when there is an obstacle node in the negative Y direction, it is not allowed to generate child nodes of Y negative X positive or Y negative X negative. There are 2 possible constraints in each direction, and a total of 16 possibilities in 4 directions. This avoids the possibility of collision between the diagonal route and obstacles in route planning, thereby ensuring that there is always a certain safe distance between the drone and the obstacle.

[0073] The route after safety planning is as follows Figure 5 This design slightly increases the length of the route, but ensures the safety of the route and eliminates the possibility of collision with obstacles.

[0074] Example 3

[0075] A method for generating autonomous inspection routes for substation drones based on the optimized A-star algorithm is proposed. The A-star algorithm and the Floyd algorithm are combined to simplify the generated optimal route. The Floyd algorithm is used to perform multiple iterative optimizations from the starting point to the target point, thus shortening the length of the planned route.

[0076] When using the conventional A-star algorithm to generate an obstacle avoidance route from the starting point to the target point, since the obstacle avoidance route can only be formed by expanding along the grid points in sequence, such a route will have redundant points and too many corner points. This paper uses Floyd's idea to simplify the route generated by the conventional A-star algorithm, delete unnecessary redundant points and corner points, shorten the length of the entire route and make the route relatively smooth, which is more conducive to the stability and smoothness of the UAV flight.

[0077] The essence of the Floyd algorithm is to use the idea of ​​dynamic programming to find the shortest route between multiple source points in a given weighted graph. This paper combines the A-star algorithm with the Floyd algorithm to simplify the generated obstacle avoidance route. By using the Floyd algorithm to perform multiple iterations from the starting point to the target point, the length of the planned obstacle avoidance route is greatly shortened. Figure 6 Shown

[0078] The route planned by the conventional A-star algorithm is A→B→C→D→E→F→G→H, that is, each step can only move along the horizontal, vertical or diagonal direction of the grid, but it is obvious that nodes C and D are redundant, that is, the route of B→C→D→E is not reasonable, there are turns and the route is long. It can be simplified to A→E→F→G→H or A→F→G→H. The route A→G→H intersects with obstacles, so it is unreasonable. Comparing the routes A→E→F→G→H or A→F→G→H, under the Floyd algorithm, the route A→F→G→H is shorter and more reasonable than the route A→E→F→G→H, but the shortest distance from the obstacle node to this route is less than 1 grid unit. Therefore, according to the route safety rules, the route A→F→G→H is abandoned and the route A→E→F→G→H is selected.

[0079] Through the Floyd algorithm optimization, some redundant nodes were deleted, the total length of the route was shortened, the route corners were reduced and relatively gentle, the route was improved, and the drone flew more smoothly and saved time.

[0080] Example 4

[0081] The method for generating autonomous inspection routes for substation drones based on the optimized A-star algorithm takes obstacle density into consideration. If the obstacle density near the current node is high, the heuristic cost is increased to prevent the drone from flying into an obstacle-dense area.

[0082] h′(n)=h(n)+w×density(n);

[0083] Among them, w is the weight coefficient, and density(n) represents the obstacle density near node n.

[0084] Example 5

[0085] A method for generating autonomous inspection routes for substation drones based on the optimized A-star algorithm. When the planned path contains unnecessary turning points and sharply changing directions, the following processing is performed:

[0086] Use B-Spline to smooth the planned path to make it smoother and more natural;

[0087]

[0088] Where C(t) is the coordinate of the curve, Pi is the control point, Ni,p(t) is the B-spline basis function of the ith control point, and its order is p;

[0089] The main characteristics of B-splines include:

[0090] The curve is controlled by the convex hull of the control points;

[0091] The shape and smoothness of the curve are determined by the locations of the control points;

[0092] The curve is a polynomial in each segment, and local adjustments can be made through control points without affecting the overall curve.

[0093] The initial path is processed using smoothing algorithms such as B-Spline. By adjusting the positions of control points and the order of B-spline basis functions, a smoother and more natural path can be obtained.

[0094] The characteristics of the B-Spline algorithm make the path a polynomial in each segment, so local adjustments can be made by adjusting the control points without affecting the overall path.

[0095] Check the redundant nodes in the path, including those on the same straight line or that can be directly connected, and remove them.

[0096] The simplified path is shorter and avoids unnecessary turning points and sudden changes in direction.

[0097] The smooth path reduces sharp turns and vibrations of the drone during flight, thereby improving flight stability.

[0098] Stable flight helps drones perform tasks better, such as filming, monitoring, etc.

[0099] The simplified path shortens the flight distance, thereby reducing the drone's flight time and energy consumption.

[0100] The optimized path enables the drone to reach its destination faster, improving flight efficiency.

[0101] The smooth and simplified path can better adapt to complex environments and obstacles, reducing the risk of collision of drones during flight.

[0102] By locally adjusting the path, flexible control of the UAV's flight trajectory can be achieved.

[0103] Path smoothing and optimization reduce the amount of calculation in the algorithm's path search process and improve the algorithm's operating efficiency.

[0104] The optimized algorithm can find the optimal path faster, thus improving the performance of the entire path planning system.

[0105] In summary, path smoothing and optimization play an important role in UAV path planning. By improving flight stability, optimizing flight efficiency, enhancing path adaptability and improving algorithm performance, it provides strong support for the widespread application of UAVs.

Claims

1. A method for generating autonomous inspection routes for substation drones based on optimized A-star algorithm, characterized in that: The nodes in the inspection route planning path are generated by the following method: the evaluation function f(n) of the node is calculated, and in each step of expanding the route node, the node with the smallest function value of the evaluation function f(n) is always selected as the child node, and then the child node is used as the parent node to expand and optimize the next step of the route child node, that is, the child node with the smallest value of the evaluation function f(n) is searched in turn to form an optimal route; The expression of the evaluation function f(n) is: f(n)=g(n)+α[h(n)+h(p)] Where: g(n) is the actual route cost from the starting node to the current node; h(n) is the estimated route cost from the current node to the target point; h(p) is the distance from the parent node of the current child node to the target point; α is the weight value.

2. According to the method of claim 1, the method is characterized in that: Put the starting point as the first node into the CLOSE list, calculate the evaluation function of up to 8 surrounding nodes. Due to flight boundaries and obstacles, the number of surrounding nodes may be less than 8, and put these nodes and node information into the OPEN list. If the current node already exists in the OPEN list, compare the evaluation function value of the node, and update the node information with the smaller evaluation function value to the OPEN list; In the OPEN list, select the node with the smallest evaluation function value as the child node of the current node; then select the selected child node as the parent node, and put its parameter information into the CLOSE list. Use the new parent node to expand the child nodes and calculate the evaluation function value of each child node. By comparing the evaluation function values, new child nodes are obtained, and the CLOSE and OPEN lists are continuously updated until the target point is found.

3. According to the method of claim 1, the method is characterized in that: When expanding child nodes, if there is an obstacle in the spatial direction of the current node, the nodes on the grid diagonal will not be expanded as child nodes, that is, no diagonal route with a distance less than 1 grid unit from the obstacle will be generated.

4. According to the method of claim 1, the method is characterized in that: g(n) is the actual route cost from the starting node to the current node; h(n) is the estimated route cost from the current node to the target point. The Euclidean distance is used as the calculation method for the route cost, and its expression is: Where: (x i ,y i ) is the current point coordinate; (x s ,y s ) is the starting point coordinate; (x t ,y t ) are the target point coordinates.

5. According to claim 3, a method for generating a substation drone autonomous inspection route based on an optimized A-star algorithm is characterized in that: Taking a two-dimensional space environment as an example, when expanding child nodes, if there are obstacles in the positive X, negative X, positive Y, and negative Y directions of the current node, the nodes on the grid diagonal will not be expanded as child nodes, that is, diagonal routes that are less than 1 grid unit away from obstacles will not be generated.

6. A method for generating a substation drone autonomous inspection route based on an optimized A-star algorithm according to claim 1 or 3, characterized in that: The generated optimal route is simplified by combining the A-star algorithm and the Floyd algorithm. The Floyd algorithm is used to perform multiple iterative optimizations from the starting point to the target point to shorten the length of the planned route.

7. The method for generating a substation drone autonomous inspection route based on the optimized A-star algorithm according to claim 1 is characterized in that: Considering the obstacle density, if the obstacle density near the current node is high, increase the heuristic cost to prevent the drone from flying into the obstacle-dense area. h′(n)=h(n)+w×density(n); Among them, w is the weight coefficient, and density(n) represents the obstacle density near node n.

8. The method for generating a substation drone autonomous inspection route based on the optimized A-star algorithm according to claim 1 is characterized in that: When the planned path contains unnecessary turning points and sharp changes in direction, the following processing is performed: Use B-Spline to smooth the planned path to make it smoother and more natural; Where C(t) is the coordinate of the curve, Pi is the control point, Ni,p(t) is the B-spline basis function of the ith control point, and its order is p; The main characteristics of B-splines include: The curve is controlled by the convex hull of the control points; The shape and smoothness of the curve are determined by the locations of the control points; The curve is a polynomial in each segment, and local adjustments can be made through control points without affecting the overall curve.

Images