Robot motion planning method based on greedy algorithm and improved hybrid A* algorithm

Abstract

The application relates to a robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm, and the method comprises the following steps: determining a traversal order of a wall-climbing robot for traversing each detection area by using the greedy algorithm, specifically comprising the following steps: selecting a detection area as a starting point; selecting a detection area closest to the current detection area and not traversed as a next detection area; adding the selected next detection area to a path and marking the detection area as visited; continuously selecting a detection area closest to the current detection area and not traversed as a next detection area until all detection areas are marked as visited; returning to the starting point; and determining the traversal order; and planning a path between adjacent detection areas by using the improved hybrid A* algorithm according to the traversal order. The method has the beneficial effect that the greedy algorithm and the improved hybrid A* algorithm are used to provide a shortest safe path that meets kinematic constraints and can be tracked for the wall-climbing robot.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robots, and in particular to a robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm. BACKGROUND

[0002] Most of the internal structures of large enclosed spaces are complex, and there are high-difficulty industrial scenes such as weak light, dust, and magnetic enclosure. Taking large boiler heating surface industrial detection as an example, at present, it mainly relies on manual operation such as hand touch and visual inspection on scaffolding, and the detection rate is high and the efficiency is low, which is difficult to achieve high coverage rate of boiler pipe detection during each boiler maintenance, which is an important reason for frequent boiler accidents. In addition, the scaffolding work under harsh conditions also often causes many personnel safety accidents and accidental damage to the boiler pipe.

[0003] The maintenance of large enclosed spaces is a complex and potentially dangerous task that requires careful planning and execution. The specific steps of large enclosed space maintenance may vary depending on the nature of the task and the characteristics of the equipment. Before performing any work, ensure compliance with local regulations and safety standards, and that all workers have received appropriate training. Safety is always the first consideration when performing maintenance on large enclosed spaces. Ensure that all steps are carefully executed to reduce potential risks and hazards.

[0004] With the help of wall climbing robot operation to replace inefficient manual detection operation in harsh environment, it can effectively reduce personnel safety accidents and accidental damage to the boiler pipe, improve system safety and reliability, ensure system safe operation, ensure safety production, and improve industrial economic benefits. However, due to the non-complete characteristics of the wall climbing robot itself and the special operation scene of the water cooling wall pipe, the wall climbing robot cannot move in the vertical direction of the current machine head direction, that is, it cannot move horizontally, and the change of its attitude angle needs to be realized through the displacement of the robot. When there are coking, manhole doors and other unreachable areas on the water cooling wall surface, the existing wall climbing robot cannot perform good obstacle avoidance control.

[0005] Therefore, how to propose a technical scheme for path planning and trajectory tracking of a wall climbing robot becomes a problem to be solved at present. SUMMARY

[0006] Technical problems to be solved

[0007] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present application provides a robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm, which solves the technical problems of path planning and trajectory tracking of a wall climbing robot.

[0008] Technical scheme

[0009] In order to achieve the above object, the main technical scheme adopted by the present application comprises:

[0010] In the first aspect, the present application provides a robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm, comprising:

[0011] The greedy algorithm is used to determine the traversal order of the wall-climbing robot in each detection area, which specifically comprises:

[0012] A detection area is selected as a starting point;

[0013] The detection area closest to the current detection area and not traversed is selected as the next detection area;

[0014] The selected next detection area is added to the path and the detection area is marked as visited, and the detection area closest to the current detection area and not traversed is continuously selected as the next detection area until all detection areas are marked as visited, the starting point is returned, and the traversal order is determined;

[0015] The improved hybrid A* algorithm is used to plan the path between adjacent detection areas according to the traversal order.

[0016] Optionally, the improved hybrid A* algorithm is used to plan the path between adjacent detection areas according to the traversal order, comprising:

[0017] The open list and the closed list are initialized, wherein the open list is used to store nodes to be traversed, and the closed list is used to store nodes that have been traversed;

[0018] The starting point is added to the open list, and is set as the current node, and it is judged whether the open list is empty; when the open list is not empty, the node with the lowest f value is selected from the open list as the current node, and the path is planned according to whether the current node is the target node; when the open list is empty, the path cannot be found; wherein f=g+h, g is the actual cost from the starting point to the current node, and h is the estimated cost from the current node to the target node.

[0019] Optionally, the path is planned according to whether the current node is the target node, comprising:

[0020] If the current node is the target node, the path is generated according to the path approximation principle, and the path generation is completed;

[0021] If the current node is not the target node, all neighbor nodes satisfying the constraint are traversed according to the constraint of the body attitude angle θ, and the path is planned according to whether the path between the current node and the neighbor node passes through an obstacle.

[0022] Optionally, the path planning according to whether the path between the current node and the neighbor node passes through the obstacle comprises:

[0023] If the path between the current node and the neighbor node passes through the obstacle, the neighbor node is deleted;

[0024] If the path between the current node and the neighbor node does not pass through the obstacle, the path planning is performed according to whether the neighbor node is in the closed list.

[0025] Optionally, the path planning according to whether the neighbor node is in the closed list comprises:

[0026] If the neighbor node is in the closed list, the node is deleted;

[0027] If the neighbor node is not in the closed list, a g value of the neighbor node is calculated, and it is determined whether the neighbor node has ever been in the open list;

[0028] If the neighbor node has ever been in the open list, the g value of the neighbor node is recalculated, and the g value of the neighbor node is determined according to the size of the current g value and the historical g value of the neighbor node;

[0029] If the neighbor node is not in the open list, an h value and an f value of the neighbor node are calculated, the neighbor node is added to the open list, and a parent node of the neighbor node is set as the current node.

[0030] Optionally, the g value of the neighbor node is determined according to the size of the current g value and the historical g value of the neighbor node, and the method comprises:

[0031] If the current g value of the neighbor node is smaller than the historical g value, the g value of the neighbor node is updated as equal to the current g value;

[0032] If the current g value of the neighbor node is greater than or equal to the historical g value, the historical g value is kept.

[0033] Optionally, the method further comprises:

[0034] Controlling the robot to perform work on each detection area according to the planned path.

[0035] Optionally, the method further comprises:

[0036] During the work of the robot, position, posture, image, speed and voltage and current information of the robot are fed back in real time;

[0037] Whether the work mode or the return is changed is determined according to the position, posture, image, speed and voltage and current information of the robot.

[0038] In a second aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed, implements the robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm described in any of the first aspects above.

[0039] Thirdly, the present invention provides a storage device, including a storage medium and a processor, wherein the storage medium stores a computer program, and when the program is executed by the processor, it implements the robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm described in any of the first aspects above.

[0040] Beneficial effects

[0041] The beneficial effects of this invention are: This invention provides a method based on a greedy algorithm and an improved hybrid A* algorithm.

[0042] The proposed robot motion planning method employs a greedy algorithm and an improved hybrid A* algorithm to plan the path for a wall-climbing robot. This method provides the robot with the shortest safe path that satisfies kinematic constraints and is trackable. Furthermore, gridding the environmental map improves the accuracy and precision of the path planning. Following the planned path reduces the probability of accidents such as slippage due to coking, manhole covers, or falls, thus improving system safety, reliability, and overall industrial economic benefits. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm provided in an embodiment of the present invention.

[0044] Figure 2 A flowchart illustrating the implementation process of another embodiment of the present invention, based on a greedy algorithm and an improved hybrid A* algorithm;

[0045] Figure 3 This is a schematic diagram illustrating a greedy algorithm for solving the sorting problem of multiple regions to be detected, provided in another embodiment of the present invention.

[0046] Figure 4 A schematic diagram of an improved hybrid A* generation provided for another embodiment of the present invention. Detailed Implementation

[0047] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] The embodiment is aimed at the working scene of the wall-climbing robot for boiler water-cooled wall maintenance in a large closed boiler space, solves the path planning and trajectory tracking problems of the wall-climbing robot operation, and for the problem, firstly, the path planning algorithm of the wall-climbing robot is explored, a planning algorithm that can meet the model and solve the limitations existing in the current field is designed. According to the known environment map, a shortest safe path that meets the kinematic constraints and can be tracked by the wall-climbing robot is planned for the wall-climbing robot, the possibility of the wall-climbing robot crashing is reduced, the operation safety is enhanced, the energy consumption of the wall-climbing robot operation is reduced, and the cost of the wall-climbing robot for maintaining the water-cooled wall is further reduced. The greedy algorithm and the improved hybrid A* algorithm can be used to design a safe path for the wall-climbing robot, which can stably track and repeat all the detection areas.

[0049] In order to better understand the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided so that the present application can be more clearly, thoroughly understood and the scope of the present application can be completely conveyed to those skilled in the art.

[0050] In a first aspect, with reference to Figure 1 The embodiment provides a robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm, comprising:

[0051] S1, a greedy algorithm is used to determine the traversal order of the wall-climbing robot for each detection area.

[0052] S1 specifically comprises:

[0053] S11, a detection area is selected as a starting point.

[0054] S12, a detection area closest to the current detection area and not traversed is selected as the next detection area.

[0055] S13, the selected next detection area is added to the path and the detection area is marked as visited, the detection area closest to the current detection area and not traversed is continuously selected as the next detection area, until all detection areas are marked as visited, the starting point is returned, and the traversal order is determined.

[0056] S2, an improved hybrid A* algorithm is used to plan a path between adjacent detection areas according to the traversal order.

[0057] In the embodiment, the starting point of the planning is the current detection area, and the target point is the next detection area. When the path is planned, the map is rasterized between the starting point and the target point to form a map node.

[0058] Optionally, the improved hybrid A* algorithm is used to plan the path between the adjacent detection areas according to the traversal order, and the improved hybrid A* algorithm comprises the following steps.

[0059] An open list and a closed list are initialized, wherein the open list is used to store nodes to be traversed, and the closed list is used to store nodes that have been traversed.

[0060] The starting point is added to the open list, and the starting point is set as the current node. It is determined whether the open list is empty. When the open list is not empty, a node with the lowest f value is selected from the open list as the current node. The path is planned according to whether the current node is the target node. When the open list is empty, the path cannot be found. Wherein, f = g + h, g is the actual cost from the starting point to the current node, and h is the estimated cost from the current node to the target node.

[0061] Optionally, the path is planned according to whether the current node is the target node, and the path planning comprises the following steps.

[0062] If the current node is the target node, the path is generated according to the path approximation principle, and the path generation is completed.

[0063] If the current node is not the target node, all neighbor nodes satisfying the constraint are traversed according to the constraint of the body attitude angle θ. The path is planned according to whether the path between the current node and the neighbor node passes through an obstacle.

[0064] Optionally, the path is planned according to whether the path between the current node and the neighbor node passes through an obstacle, and the path planning comprises the following steps.

[0065] If the path between the current node and the neighbor node passes through an obstacle, the neighbor node is deleted.

[0066] If the path between the current node and the neighbor node does not pass through an obstacle, the path is planned according to whether the neighbor node is in the closed list.

[0067] Optionally, the path is planned according to whether the neighbor node is in the closed list, and the path planning comprises the following steps.

[0068] If the neighbor node is in the closed list, the neighbor node is deleted.

[0069] If the neighbor node is not in the closed list, a g value of the neighbor node is calculated, and it is determined whether the neighbor node has ever been in the open list;

[0070] If the neighbor node has ever been in the open list, the g value of the neighbor node is recalculated, and the g value of the neighbor node is determined according to the size of the current g value and the historical g value of the neighbor node;

[0071] If the neighbor node is not in the open list, an h value and an f value of the neighbor node are calculated, the neighbor node is added to the open list, and a parent node of the neighbor node is set as the current node.

[0072] Optionally, the g value of the neighbor node is determined according to the size of the current g value and the historical g value of the neighbor node, and the method comprises the following steps of:

[0073] If the current g value of the neighbor node is smaller than the historical g value, the g value of the neighbor node is updated as equal to the current g value;

[0074] If the current g value of the neighbor node is greater than or equal to the historical g value, the historical g value is kept.

[0075] Optionally, the method further comprises the following steps of:

[0076] The robot is controlled to perform work on each detection area according to the planned path.

[0077] Optionally, the method further comprises the following steps of:

[0078] During the work of the robot, position, posture, image, speed and voltage and current information of the robot are fed back in real time;

[0079] Whether the work mode or the return is changed is determined according to the position, posture, image, speed and voltage and current information of the robot.

[0080] The robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm can provide a shortest safe path meeting kinematic constraints and traceable by the wall-climbing robot for the wall-climbing robot, and the grid processing of the environment map can improve the accuracy and accuracy of the wall-climbing robot path planning. The wall-climbing robot performs work according to the planning result, which can reduce the probability of accidents such as injury caused by sliding and falling of the wall-climbing robot due to coking and manhole doors during the travel process, effectively reduce personnel safety accidents, improve system safety and reliability, ensure safe operation of the system, ensure safety production, and improve industrial economic benefits.

[0081] The robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm will be further described below in combination with a specific embodiment:

[0082] First, for multiple regions to be detected, they are sorted to find a sequence that can traverse all the regions to be detected with the smallest cost. This problem can be abstracted as a traveling salesman problem, where the wall-climbing robot is considered as a "salesman" and each key detection region is considered as an "city" that the salesman needs to visit. The goal is to find the best traversal order for the wall-climbing robot to traverse each key detection region without missing any and return to the starting point.

[0083] There are many methods to solve the traveling salesman problem (TSP), some of which include:

[0084] Brute Force: This is an intuitive but computationally expensive method. It tries to list all possible paths and calculates their total length, and finally selects the shortest path. However, as the number of cities increases, the computational complexity of the brute force method grows exponentially, so it is not practical for large-scale problems.

[0085] Dynamic Programming: TSP can be solved by dynamic programming method. Dynamic programming algorithm uses a bottom-up approach to calculate the optimal solution of sub-problems step by step, and finally gets the solution of the whole problem. Although dynamic programming can solve TSP in theory, it still has high computational complexity for large-scale problems.

[0086] Greedy Algorithm: Greedy algorithm is a method of local optimal selection. In each step, the optimal solution in the current state is selected to gradually build the path. Greedy algorithm cannot guarantee to find the global optimal solution, but it can provide a good approximate solution in some cases.

[0087] Approximation Algorithm: Approximation algorithm is a class of algorithms that provide approximate solutions to problems within polynomial time. For example, minimum spanning tree algorithms such as Prim's algorithm and Kruskal's algorithm can be used for approximate solution of TSP.

[0088] Genetic Algorithm: Genetic algorithm is a heuristic algorithm that simulates natural selection and genetic mechanisms to search the solution space. It can be used to solve TSP, especially when dealing with large-scale problems.

[0089] Ant Colony Algorithm: Ant colony algorithm simulates the behavior of ants in the process of finding food. Ant colony algorithm is widely used in TSP, and ants update the path through the deposition and evaporation of pheromone in the process of searching the path.

[0090] Each of these methods has its own advantages and disadvantages, and is suitable for TSP problems of different sizes and characteristics. For small-scale problems, the exhaustive method and dynamic programming method may be feasible; for large-scale problems, heuristic methods such as greedy algorithm, approximation algorithm, genetic algorithm and ant colony algorithm may be more practical. The choice of the appropriate method depends on the size and characteristics of the problem and the accuracy requirements of the solution. Our application scenario requires fast calculation speed and simple algorithm implementation, although the greedy algorithm cannot guarantee to find the optimal solution for TSP, but in some cases it can provide a better approximate solution with lower time complexity, so the greedy algorithm is chosen for solving.

[0091] After sorting the multiple key detection areas, path planning between adjacent key detection areas is needed. Path planning refers to the process of finding the best path from the starting point to the target point in a given environment. This problem covers many different fields, including robot navigation, autonomous driving, logistics management, etc. Here are some common path planning methods:

[0092] Shortest path algorithm: Dijkstra's algorithm finds the shortest path from the starting point to the target point through the shortest path between the nodes of the graph; A* algorithm combines Dijkstra's algorithm and heuristic search, estimates the distance to the target to speed up the search process.

[0093] Graph-based search algorithm: Depth-first search (DFS) is to traverse along the depth of the graph until the target is found or the maximum depth is reached; Breadth-first search (BFS) is to expand the search layer by layer according to the distance, ensuring that the nodes close to the starting point are explored first.

[0094] Sampling-based search algorithm: Rapidly-exploring Random Trees (RRT) generates a path by randomly sampling points in the configuration space and using fast exploration to expand the tree structure; PRM (Probabilistic Roadmap) samples a large number of random points in the configuration space, establishes a network, and then uses graph search algorithm to find the path.

[0095] Optimization-based algorithm: Genetic algorithm optimizes the quality of the path by simulating natural selection and genetic mechanisms; Simulated Annealing algorithm simulates the process of metal smelting, gradually reduces the system temperature to find the global optimal solution.

[0096] Artificial neural network: Deep reinforcement learning uses neural networks and reinforcement learning techniques to let the system learn the best path through trial and error; Model Predictive Control is to predict the future state of the environment and optimize the control input to plan the path.

[0097] Hierarchical planning is to divide the planning task into multiple levels, which is convenient for handling large-scale and complex environments.

[0098] Hybrid methods combine different methods to overcome the limitations of individual methods and improve the performance of path planning. The choice of path planning method usually depends on the nature of the specific problem, including the complexity of the environment, path constraints, real-time requirements, etc. In practical applications, it is often necessary to combine multiple methods to better adapt to different scenarios and requirements.

[0099] The current obstacle avoidance control model for moving bodies is the integrator model, which is commonly used in control systems to handle complete constraints. However, it cannot adapt to the non-holonomic characteristics of wall-climbing robots. Therefore, we need to design a trackable path planning algorithm that meets non-holonomic constraints. The typical method to solve this problem is the hybrid A* algorithm.

[0100] Classic A* only considers the x and y dimensions, while hybrid A* introduces the theta dimension. In classic A*, the car is treated as a point, while in hybrid A*, theta is added to represent the orientation, with x and y representing the center of the rear axle of the vehicle.

[0101] In this way, the search space changes from a two-dimensional table to an x-y-θ state space.

[0102] In hybrid A*, it can be understood as expanding outward from the initial small line segment on a chessboard grid. Each time a small curve can be expanded, and the node can be at any position on the chessboard grid until the small line segment connecting the end point is connected.

[0103] The classic A* algorithm treats each state grid as a pixel point, i.e., the robot path is composed of lines connecting the centers of each square. The Field D* algorithm allows the path to land on the edges of the square, but the path is still connected by straight line segments. The Hybrid A* algorithm allows the path to land on the state grid, and the connecting lines are replaced by certain curves, which meet the non-holonomic constraints.

[0104] In classic A*, the way to expand nodes is to find adjacent nodes. However, since hybrid A* considers vehicle dynamics and adds the theta dimension, hybrid A* cannot simply consider adjacent nodes as expansion targets. Its expansion needs to consider vehicle dynamics.

[0105] The Reeds-Shepp curve generation mechanism is another feature of the hybrid A*. The Reeds-Shepp curve can connect the start and end poses in a manner that basically meets the vehicle dynamics constraints, but does not consider the obstacle avoidance factor at all. Another commonly used curve is the dubins curve. The RS curve and the dubins curve are similar, both of which are circular plus straight line paths. The difference is that the RS curve allows the vehicle to reverse.

[0106] The Reeds-Shepp curve is composed of only three parts, a left circular arc, a right circular arc, and a straight line. Corresponding to the non-holonomic model, it is a left turn, a right turn, and a straight line. Therefore, the path is some circular arcs and straight lines that are subject to the non-holonomic model turn radius constraints.

[0107] Although the Reeds-Shepp curve does not consider the obstacle avoidance factor, because it generates at a speed fast enough, it can first generate a path that meets the dynamics and then check whether the path meets the collision condition. Specifically, during the execution of the hybrid A*, the Reeds-Shepp curve can be triggered intermittently to generate a curve that connects the current pose and the end pose, and then it is determined whether it meets the collision condition. If it meets the collision condition, the problem is solved. If it does not meet the collision condition, the result is discarded, and the normal search logic continues to be executed.

[0108] This way of adding the Reeds-Shepp curve to the traditional A* search greatly improves the search speed of the hybrid A*. Because the end pose can also be met, it is commonly used in non-holonomic model path planning.

[0109] However, in the present application, the hybrid A* algorithm is improved. Because the non-holonomic constraint model in the technical field is the wall-climbing robot, it not only has non-holonomic constraints, but also has other motion constraints and attitude constraints.

[0110] The motion constraint is caused by the mechanical structure of the wall-climbing robot. The wall-climbing robot body is connected to many cables, which are located below the wall-climbing robot body. When the wall-climbing robot is working autonomously, if the head moves downward, the cables will have a certain probability to hinder or even interfere with the work of the wall-climbing robot. Therefore, the mechanical structure causes the wall-climbing robot to be unable to move the head downward when working.

[0111] The attitude constraint is because when the wall-climbing robot moves in the direction perpendicular or approximately perpendicular to the water cooling wall pipe, the irregular and not smooth boiler water cooling wall will cause the wall-climbing robot to vibrate, causing damage to the wheels and the balance instrument loaded inside. Therefore, the attitude of the wall-climbing robot needs to be constrained to avoid the wall-climbing robot moving in the direction perpendicular to the water cooling wall pipe, that is, the range of the angle between the head of the wall-climbing robot and the vertical direction (attitude angle) is [-45°, 45°].

[0112] According to the above constraints, the path planning algorithm for wall-climbing robots with such constraints is improved, that is, the hybrid A* algorithm is improved, and new methods are proposed for neighbor node selection and path trajectory fitting. Case experiments and analysis are conducted.

[0113] The greedy algorithm in solving the Traveling Salesman Problem (TSP) is based on local optimal selection, gradually constructing the path until the end condition of the problem is met. TSP is a classic combinatorial optimization problem, aiming to find a path that makes the traveling salesman visit each city once and only once, finally return to the starting city, and minimize the total path length.

[0114] Here is a basic idea of using the greedy algorithm to solve TSP:

[0115] 1. Select the starting city: Choose a city as the starting point.

[0116] 2. Greedy selection: In the current state, choose the city closest to the current city and not yet visited as the next city to visit.

[0117] 3. Update state: Add the selected city to the path and mark it as visited.

[0118] 4. Repeat greedy selection: Repeat the above process until all cities have been visited.

[0119] 5. Return to the starting city: Connect the last city in the path to the starting city to form a closed path.

[0120] 6. End condition: Meet the end condition of the problem, such as all cities being visited.

[0121] 7. Optimize the path: After the end, some optimization of the path may be needed to ensure that the path obtained is the shortest path.

[0122] The idea of this greedy algorithm is to choose the optimal solution in the current state at each step, gradually constructing the path.

[0123] The improved hybrid A* algorithm is a heuristic search algorithm used to find the shortest path from the starting point to the ending point. It is based on the evaluation function (heuristic function) to evaluate the priority of each node, and selects the node with the highest priority to expand until the ending point is found or all reachable nodes are expanded.

[0124] The steps to generate a path using the improved hybrid A* algorithm are as follows:

[0125] 1. Initialize the open list and closed list.

[0126] The open list is used to store nodes to be explored, and the closed list is used to store nodes that have been explored.

[0127] 2. Add the start node to the open list and set it as the current node. Determine whether the open list is empty:

[0128] a. When the open list is not empty, perform the following steps:

[0129] (1) Select the node with the lowest f value (f = g + h, where g is the actual cost from the start node to the current node, and h is the estimated cost from the current node to the target node) from the open list as the current node.

[0130] (2) Determine whether the current node is the target node:

[0131] (a) If the current node is the target node, generate the path according to the path approximation principle, and the path generation is complete.

[0132] (b) If the current node is not the target node, traverse all neighbor nodes that satisfy the constraint according to the constraint of the body attitude angle θ, and determine whether the path between the current node and the neighbor node passes through an obstacle:

[0133] (a) If it passes through an obstacle, delete the neighbor node.

[0134] (b) If it does not pass through an obstacle, determine whether it is in the closed list:

[0135] (a) If it is in the closed list, delete the node.

[0136] (b) If it is not in the closed list, calculate the g value of the neighbor node, and determine whether the neighbor node has ever been in the open list:

[0137] (a) If it is in the open list, recalculate the g value of the neighbor node, and compare the current g value and the historical g value of the neighbor node:

[0138] (a) If the current g value < historical g value, update the g value of the neighbor node to g = current g value.

[0139] (b) If the current g value ≥ historical g value, keep the historical g value

[0140] (b) If it is not in the open list, calculate the h value and f value of the neighbor node, add the neighbor node to the open list, and set the parent node of the neighbor node as the current node.

[0141] b. If the open list is empty, it means that the path cannot be found.

[0142] When the path generation is completed, the final path can be obtained by backtracking from the target node to the starting node. In this process, the role of the evaluation function is to evaluate the distance of each node to the end point, which is used to select the node with the highest priority for expansion.

[0143] Figure 2 The case implementation process diagram is based on the greedy algorithm and the improved hybrid A* algorithm. According to the environmental information, the wall-climbing robot information, the to-be-detected area information, and the obstacle information (unreachable area information) provided by the system, the best detection order is generated according to the greedy algorithm, and the detection order of a series of to-be-detected areas is given. Then, the improved hybrid A* algorithm is used for path planning between two adjacent to-be-detected areas, and a traceable, minimum-cost safe path is generated.

[0144] The task flow of the wall-climbing robot is as follows: the system cooperates to issue the motion trajectory of the wall-climbing robot generated by the algorithm to the wall-climbing robot. The wall-climbing robot starts from the hovering point and climbs to the key detection area (target climbing point, hereinafter referred to as target detection area). After reaching the key detection area, the wall-climbing robot performs dust removal, thickness measurement, and other operations in the target detection area. In this process, the wall-climbing robot needs to feed back the current position, attitude, image, speed, voltage, current, and other information to the cooperative system in real time. The cooperative system determines whether to change the operation mode (manual / automatic) or whether to return. If the operation mode needs to be changed or the wall-climbing robot needs to return, the wall-climbing robot operates according to the new operation mode or returns. If the operation mode does not need to be changed or the wall-climbing robot does not need to return, the wall-climbing robot climbs to the next key detection area and starts a new round of key detection area operation.

[0145] Figure 3 The diagram shows that the greedy algorithm can give a relatively optimal traversal order to a certain extent, and the effect is good.

[0146] Figure 4 The simulation diagram of the improved hybrid A* algorithm for wall-climbing robot path planning. The gray box area in the diagram is the unreachable area in the environment map. The star marks the starting and ending point positions of the wall-climbing robot path. The dashed line is the ordinary path planning result generated by the A* algorithm, and the solid line is the optimized path planning result obtained by the improved hybrid A* algorithm.

[0147] The algorithm introduces the motion constraints caused by the nonholonomic characteristics of the wall-climbing robot, the attitude constraints in the actual operation process and the position constraints caused by the unreachable area in the environment on the basis of the classical A* algorithm, considers multiple constraints, uses a curved path instead of the original polyline path, extends the classical A* algorithm, and aims to generate a shortest safe path meeting the kinematic constraints and being trackable by the wall-climbing robot.

[0148] 1. The planning result guarantees the kinematic constraints caused by the nonholonomic characteristics of the wall-climbing robot, and guarantees the trackability of the wall-climbing robot to the planning result.

[0149] 2. The output path in the planning result is prevented from overlapping with the unreachable area in the environment map, and the safety of the wall-climbing robot in the operation process is guaranteed.

[0150] The wall-climbing robot is subjected to path planning, a shortest safe path meeting the kinematic constraints and being trackable by the wall-climbing robot can be provided for the wall-climbing robot, meanwhile, the grid processing of the environment map can improve the accuracy and accuracy of the wall-climbing robot path planning. The wall-climbing robot can reduce the probability of events such as injury caused by sliding and falling of the wall-climbing robot due to coking and manhole doors in the advancing process, effectively reduce personnel safety accidents, improve system safety, reliability, guarantee system safe operation, guarantee safety production, and improve industrial economic benefits.

[0151] In a second aspect, an embodiment of the present application provides a computer readable storage medium having a computer program stored thereon, the program being executed to implement the robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm in any one of the first aspect.

[0152] In a third aspect, an embodiment of the present application provides a storage device including a storage medium and a processor, the storage medium storing a computer program, and the program being executed by the processor to implement the robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm in any one of the first aspect.

[0153] Those skilled in the art should understand that embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0154] It is apparent that those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, the present application should be construed to encompass all such modifications and variations as fall within the scope of the present claims and their equivalents.

[0155] Although embodiments of the present application have been shown and described above, it is to be understood that the above examples are meant to be illustrative only and the present application is not to be limited to the above examples but can be modified, altered, replaced and varied by those skilled in the art within the scope of the present application.

Claims

1. A robot motion planning method based on a greedy algorithm and an improved hybrid A* algorithm, characterized in that, include: A greedy algorithm is used to determine the traversal order of each detection area by the wall-climbing robot, which specifically includes: Select a detection area as the starting point; Select the detection region that is closest to the current detection region and has not been traversed as the next detection region; Add the selected next detection region to the path and mark it as visited. Continue to select the detection region that is closest to the current detection region and has not been traversed as the next detection region, until all detection regions are marked as visited. Return to the starting point, and the traversal order is determined. Based on the traversal order, an improved hybrid A* algorithm is used to plan paths between adjacent detection regions; The step of using the improved hybrid A* algorithm to plan paths between adjacent detection regions according to the traversal order includes: Initialize the open list and the closed list, where the open list is used to store the nodes to be traversed and the closed list is used to store the nodes that have been traversed; Add the starting point to the open list and set it as the current node. Check if the open list is empty. If the open list is not empty, select the node with the lowest f value from the open list as the current node. Perform path planning based on whether the current node is the target node. If the open list is empty, no path can be found. Here, f = g + h, where g is the actual cost from the starting point to the current node and h is the estimated cost from the current node to the target node. The path planning based on whether the current node is the target node includes: If the current node is the target node, then a path is generated according to the path approximation principle, and the path generation is complete; If the current node is not the target node, based on the constraint of the body attitude angle θ, traverse all neighboring nodes that satisfy the constraint, and perform path planning based on whether the path between the current node and neighboring nodes passes through obstacles.

2. The robot motion planning method based on greedy algorithm and improved hybrid A* algorithm according to claim 1, characterized in that, The path planning based on whether the path between the current node and its neighboring nodes passes through obstacles includes: If the path between the current node and its neighbors passes through an obstacle, then delete that neighbor node. If the path between the current node and its neighboring nodes does not pass through any obstacles, path planning is performed based on whether the neighboring nodes are in the closed list.

3. The robot motion planning method based on greedy algorithm and improved hybrid A* algorithm according to claim 2, characterized in that, The path planning based on whether the neighbor node is in the closed list includes: If a neighboring node is in the closed list, delete that node. If a neighboring node is not in the closed list, calculate the g value of that neighboring node to determine whether it was ever in the open list. If the neighboring node was previously in the open list, recalculate the neighboring node's g value, and determine the neighboring node's g value based on the current and historical g values. If the neighboring node is not in the open list, calculate the h and f values ​​of the neighboring node, add the neighboring node to the open list, and set the parent node of the neighboring node to the current node.

4. The robot motion planning method based on greedy algorithm and improved hybrid A* algorithm according to claim 3, characterized in that, The step of determining the neighbor node's g value based on the neighbor node's current g value and historical g value includes: If the current g value of a neighboring node is less than its historical g value, update the g value of that neighboring node to the current g value. If the current g value of a neighboring node is greater than or equal to its historical g value, maintain the historical g value.

5. The robot motion planning method based on greedy algorithm and improved hybrid A* algorithm according to claim 1, characterized in that, The method further includes: The robot is controlled to perform its work in each inspection area according to the planned path.

6. The robot motion planning method based on greedy algorithm and improved hybrid A* algorithm according to claim 1, characterized in that, The method further includes: During robot operation, the robot's position, posture, image, speed, and voltage and current information are fed back in real time. The system determines whether to change the operating mode or return to base based on the robot's position, posture, image, speed, and voltage / current information.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm as described in any one of claims 1 to 6.

8. A storage device comprising a storage medium and a processor, wherein the storage medium stores a computer program, characterized in that, When the processor executes the computer program, it implements the robot motion planning method based on the greedy algorithm and the improved hybrid A* algorithm as described in any one of claims 1 to 6.

Images