Path planning method integrating A* and improved RRT-Connect
By integrating A* and improving RRT-Connect algorithm, path planning is optimized, and the problems of low path planning efficiency and insufficient dynamic obstacle avoidance capabilities in the existing technology are solved, and fast feasible paths are achieved in high-precision maps and dynamic obstacle environments are realized. Planning and dynamic obstacle avoidance.
Patent Information
- Application Number
- CN202210326633.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-30
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-03-30
AI Technical Summary
The existing path planning method is low efficiency and time-consuming in high-precision maps and dynamic obstacle environments, making it difficult to achieve dynamic obstacle avoidance. The search-based method is slow, and the sampling-based method is easy to fall into local optimization. It is based on mathematical models and neural networks. The method is difficult to implement.
Fusion of A* and improved RRT-Connect algorithm, optimize path planning through heuristic sampling and adaptive adjustment of iteration times, and combining the A* algorithm to accurately plan paths under local complex structures to achieve dynamic obstacle avoidance.
It improves the efficiency and speed of path planning, avoids the long time-consuming and blind growth path costs of path planning, and realizes fast feasible path planning and dynamic obstacle avoidance in a dynamic environment.
Smart Images

Figure CN114812555B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of path planning, and in particular relates to a path planning method, a storage medium and a computer device integrating A* and improved RRT-Connect. Background Art
[0002] Path planning is one of the key technologies for robot autonomous navigation. Existing robot path planning methods include search-based, sampling-based, mathematical model-based, and neural network-based methods.
[0003] However, the path planning algorithms based on mathematical models and neural networks are only discussed at the theoretical level in the industry, and are difficult to implement in practice; the search-based path planning is slow, and the path planning time is seriously prolonged when the map accuracy and area increase. When there are active obstacles, the path may continue to fail or even navigation may fail; the sampling-based path planning is too random, and the resulting path cost is high, which makes it easy to fall into the dilemma of local optimality and difficult to achieve global optimality; the existing common path planning methods are all based on static maps, and it is difficult to dynamically avoid obstacles in the actual environment corresponding to the map. Summary of the invention
[0004] In order to overcome one or more defects and shortcomings of the prior art, the purpose of the present invention is to provide a path planning method that integrates A* and improved RRT-Connect, so as to improve the efficiency of path planning and reduce time consumption, avoid the problem of blindly growing path costs caused by sampling-based path planning methods, and adapt to dynamic changes in the map environment.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions:
[0006] A path planning method integrating A* and improved RRT-Connect includes the following steps:
[0007] Initialization: load the map, set the number of iterations, iteration time, starting point, and target point;
[0008] Plan the path according to the improved RRT-Connect algorithm: Use the improved RRT-Connect algorithm to set heuristic sampling points in the map for random sampling, perform tree generation operations on the starting point tree and the target point tree to expand the corresponding nearest points, and then backtrack from the latest nearest point to the root node of the starting point tree and the root node of the target point tree, and determine whether the starting point tree and the target point tree converge within the set number of iterations or iteration time; if so, directly splice the backtracked path to obtain the map path from the starting point to the target point; if not, obtain two paths, and then perform path search according to the A* algorithm; the heuristic sampling points are set to use the starting point tree and the target point tree as the sampling area with a set probability, and use the map area outside the starting point tree and the target point tree as the sampling area with a set probability;
[0009] Path search based on A* algorithm: Use A* algorithm to search for the local feasible path with the minimum path cost from the starting point tree to the target point tree on the nearest points corresponding to the starting point tree and the target point tree;
[0010] Three-segment path splicing: Splice the two paths obtained by planning the path according to the improved RRT-Connect algorithm and the local feasible path obtained by path search according to the A* algorithm to obtain the map path from the starting point to the target point.
[0011] Preferably, the initialization further comprises searching a path from a starting point to a target point using an A* algorithm when performing path planning for the first time, and using the obtained map path as a reference for a map path obtained by splicing the three path segments.
[0012] Preferably, the steps of planning a path according to the improved RRT-Connect algorithm are as follows:
[0013] Set the starting point and expansion step size for the starting point tree and the target point tree respectively, and then use the heuristic sampling point setting to randomly generate sampling points on the map;
[0014] Searching for the nearest neighbor points of the starting point tree and the target point tree respectively, wherein the nearest neighbor points are the sampling points of the starting point tree or the target point tree that are closest to each other;
[0015] Starting from the nearest point, set the starting point tree and the target point tree to grow temporary nodes in the direction of the sampling point with an extended step length, and establish a temporary connection between the nearest point and the temporary node;
[0016] Whether the temporarily connected line falls into the obstacle range of the map or exceeds the map boundary; if so, disconnect the temporary connection; if not, add the temporary node to the starting point tree or the target point tree accordingly;
[0017] Search for the optimal parent node of the temporary node to update the starting point tree or the target point tree accordingly;
[0018] Determine whether the starting point tree and the target point tree are merged to output the map path.
[0019] Further, the searching for the optimal parent node of the temporary node to update the starting point tree or the target point tree accordingly specifically includes:
[0020] Calculate the sampling point with the smallest path cost to the temporary node and select it as the candidate for the optimal parent node; establish the connection between the candidate optimal parent node and the temporary node and determine whether the connection falls into the obstacle range in the map or exceeds the map boundary;
[0021] If yes, disconnect the connection between the candidate optimal parent node and the temporary node;
[0022] If not, it is taken as the optimal parent node and merged into the starting point tree or the target point tree accordingly.
[0023] Further, the determining whether the starting point tree and the target point tree merge specifically includes:
[0024] If they converge, the latest temporary nodes of the starting point tree and the target point tree are used to trace back to the starting point and the target point respectively, and then the two traced paths are spliced to obtain the map path from the starting point to the target point;
[0025] If there is no convergence, determine whether the number of iterations or iteration time exceeds the set value during initialization;
[0026] If exceeded, restart the sampling point setting;
[0027] If not, then the step of performing path search according to the A* algorithm is performed.
[0028] Furthermore, the steps of performing path search according to the A* algorithm are as follows:
[0029] Mark the latest nearest point in the starting point tree as the A* starting point, and the latest nearest point in the target point tree as the A* end point, then calculate the path cost of all paths from the A* starting point to the A* end point; take the path with the smallest path cost as the local feasible path from the A* starting point to the A* end point.
[0030] Furthermore, the calculation of the path cost of all paths from the A* starting point to the A* end point specifically includes:
[0031] Initialize the open and closed containers in the A* algorithm;
[0032] Mark the cumulative cost of the starting point of A* as zero, calculate the estimated cost between the two nodes, add the cumulative cost and the estimated cost to get the path cost, and then put the path cost into the open container;
[0033] Determine whether the open container is empty; if the open container is empty, the path planning fails, and the path cost is recalculated by selecting a new adjacent node; if the open container is not empty, the node with the smallest path cost is taken from the open container as the current node, and then the current node is added to the closed container;
[0034] Determine whether the current node is the A* end point; if the current node is the A* end point, output a local feasible path, and then perform the three-segment path splicing; if the current node is not the A* end point, traverse the adjacent nodes of the current node to continue the path search.
[0035] Furthermore, the traversing of the adjacent nodes of the current node and continuing the path search specifically includes:
[0036] Calculate the path cost from the current node to one of the adjacent nodes;
[0037] Determine whether the adjacent node is in the open container; if not, compare the path cost from the A* starting point to the adjacent node with the minimum path cost in the current open container; if the path cost from the A* starting point to the adjacent node is smaller, update the open container, take the latest path cost corresponding to the adjacent node as the minimum path cost in the open container, and add the adjacent node as the new current node to the closed container; if it is, put the adjacent node and its path cost into the open container;
[0038] Determine whether all adjacent nodes of the current node have been traversed; if all adjacent nodes of the current node have been traversed, return to the step of determining whether the open container is empty; if all adjacent nodes of the current node have not been traversed, return to the step of calculating the path cost from the current node to its next adjacent node.
[0039] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:
[0040] By adopting heuristic ideas to optimize the random sampling process based on the RRT-Connect algorithm, the improved RRT-Connect can quickly plan feasible paths in a dynamic environment where obstacles are constantly moving. Combined with the advantage of the A* algorithm in accurately planning the optimal path under local complex structures, the algorithm is switched to adapt to different scenarios by using an adaptive adjustment method of the number of iterations, which effectively improves the search speed and shortens the planning time, avoiding the problems of low search efficiency and long search time when using a search-based path planning method alone, avoiding continuous path failure, and blindly growing and expanding path costs when using a sampling-based path planning method alone; by utilizing the randomness of setting sampling points, the path can be re-searched immediately after the path planning fails due to changes in obstacles in the map, thus realizing the function of dynamic obstacle avoidance. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 It is a schematic diagram of the overall process of a path planning method integrating A* and improved RRT-Connect of the present invention;
[0042] Figure 2 for Figure 1 Schematic diagram of the process of path planning using improved RRT-Connect;
[0043] Figure 3 for Figure 1 Schematic diagram of the process of using A* for path planning;
[0044] Figure 4 for Figure 1 Schematic diagram of the path splicing of A* and improved RRT-Connect algorithm;
[0045] Figure 5 for Figure 1 Schematic diagram of the algorithm performance of the fusion A* and improved RRT-Connect.
[0046] Figure 6 For execution Figure 1 A map of the experimental field where the dynamic obstacle avoidance method is performed;
[0047] Figure 7 (a) Figure 1 The motion trajectory of the robot at the beginning of the dynamic obstacle avoidance experiment using the method and the A* algorithm;
[0048] Figure 7 (b) Figure 1 The motion trajectory of the robot in the dynamic obstacle avoidance experiment at 4.2 seconds using the middle method and the A* algorithm;
[0049] Figure 7 (c) Figure 1The robot's motion trajectory diagram of the method and the A* algorithm in the dynamic obstacle avoidance experiment at 8.4 seconds;
[0050] Figure 7 (d) Figure 1 The motion trajectory of the robot in the dynamic obstacle avoidance experiment at 12.6 seconds using the method and the A* algorithm;
[0051] Figure 7 (e) Figure 1 The motion trajectory of the robot in the dynamic obstacle avoidance experiment at 16.8 seconds using the middle method and the A* algorithm;
[0052] Figure 7 (f) Figure 1 The motion trajectory of the robot in the dynamic obstacle avoidance experiment at 22 seconds using the method and the A* algorithm; DETAILED DESCRIPTION
[0053] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments thereof. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0054] Example
[0055] like Figure 1 As shown, a path planning method integrating A* and improved RRT-Connect in this embodiment includes the following steps in sequence.
[0056] S1. Initialization, including:
[0057] S11, set the number of iterations to start from 100, the maximum number of iterations to 800, and the iteration time to 3 seconds, and the robot loads the latest map containing dynamic obstacles;
[0058] At the same time, the content of collision detection is defined as: detecting whether the line between two points falls into the range of the obstacle area in the map or exceeds the map boundary, or detecting whether the node falls into the range of the obstacle area in the map or exceeds the map boundary. If it falls into the range of the obstacle area or exceeds the map boundary, it is considered that the line or node has collided with the map, otherwise it is considered that no collision has occurred;
[0059] S12, the robot identifies the channels and channel widths existing in the map and obtains its own position in the map, and then sets the starting point and the target point;
[0060] S13, performing path planning to obtain a reference map path. The first path planning directly uses the A* algorithm, and the subsequent path planning is updated by the map path returned in step S5;
[0061] S2, such as Figure 2 As shown in the figure, the spanning tree is extended according to the improved RRT-Connect algorithm. The spanning tree includes a starting point tree and a target point tree, and the two parts start to grow synchronously, including:
[0062] S21, set the starting point and expansion step of the spanning tree: take the starting point as the root node of the starting point tree for path growth, take the target point as the root node of the target point tree for path growth, and set the expansion step used in the spanning tree process to 0.1 meters;
[0063] S22, performing heuristic setting of sampling points: setting the sampling area on the starting point tree or the target point tree with a probability of 20% and using the starting point or the target point as the sampling point, and setting the sampling area in an area outside the map generation tree with a probability of 80%;
[0064] S23, search for the nearest neighbor: after setting the sampling probability in step S22, execute the random sampling function to generate sampling points, and then obtain the current starting point tree or target point tree respectively (the starting point tree at the first sampling only contains the starting point, and the target point tree only contains the target point), set the node closest to the sampling point corresponding to the starting point tree as the nearest neighbor of the starting point tree, and set the node closest to the sampling point corresponding to the target point tree as the nearest neighbor of the target point tree;
[0065] S24, respectively set corresponding temporary nodes for the starting point tree and the target point tree: starting from the nearest point, grow the temporary node in a straight line in the direction of the corresponding sampling point with an expansion step of 0.1 meters, and then connect the nearest point and the temporary node to establish a temporary connection between the nearest point and the temporary node;
[0066] S25. Perform collision detection on temporarily connected lines:
[0067] Match the temporarily connected lines with the map, and detect whether the temporarily connected lines collide with the map;
[0068] If the temporarily connected line collides with the map, the temporary connection is disconnected and the temporary node is not added to the starting point tree or the target point tree;
[0069] If the temporarily connected line does not collide with the map, confirm that the temporary node is added to the starting point tree or the target point tree accordingly, and then execute step S26;
[0070] S26, search for the optimal parent node: calculate the path cost from the root node through the nearest point or the sampling point within a radius of 0.1 meters of the temporary node and located on the spanning tree to the temporary node, and then select the node with the smallest path cost as the candidate for the optimal parent node;
[0071] Establish a connection between the candidate optimal parent node and the temporary node and perform collision detection;
[0072] If the connection between the candidate optimal parent node and the temporary node collides with the map, the connection between the candidate optimal parent node and the temporary node is disconnected; if the connection between the candidate optimal parent node and the temporary node does not collide with the map, it is used as the optimal parent node and the connection is maintained, so that the temporary node is added to the starting point tree or the target point tree accordingly, thereby updating the starting point tree or the target point tree;
[0073] S27, determining whether the starting point tree and the target point tree converge;
[0074] If they have already converged, the latest temporary node of the spanning tree is used to trace back to the starting point and the target point respectively, and then the two paths are spliced to obtain the planned map path from the starting point to the target point, and then step S5 is executed;
[0075] If there is no convergence, execute step S28;
[0076] S28, determining whether the threshold of the number of iterations exceeds 800;
[0077] If the number of iterations exceeds the threshold, the path planning fails, and the process returns to step S22 to reset the sampling points.
[0078] If the number of iterations does not exceed the threshold, determine whether the iteration time exceeds 3 seconds;
[0079] If the iteration time exceeds 3 seconds, the path planning fails, and the process returns to step S22 to reset the sampling points.
[0080] If the iteration time does not exceed 3 seconds, execute step S3;
[0081] S3, such as Figure 3 As shown, the path search is performed according to the A* algorithm, specifically including:
[0082] S31. Calculate the path cost from the current node to its adjacent node in the process from the starting point of A* to the end point of A*:
[0083] Mark the latest nearest point in the starting point tree as the A* starting point, and the latest nearest point in the target point tree as the A* end point, and then initialize the open container open_set and the closed container close_list in the A* algorithm;
[0084] The cumulative cost g(n) of the starting point of A* is marked as zero (that is, the starting point of A* is the first current node), and the estimated cost h(n) between the two nodes is calculated by the Euclidean distance formula. The calculation formula of the estimated cost h(n) is as follows:
[0085]
[0086] Among them, (x n ,y n ) represents the map coordinates of the current node, (x goal ,y goal ) represents the map coordinates of the adjacent nodes, B is the weight for calculating the estimated cost, B is a constant, subscript n represents the current node, subscript goal represents the adjacent nodes of the current node, n and goal are both positive integers, and the specific values are used to represent the meaning of the first node;
[0087] Add the cumulative cost g(n) and the estimated cost h(n) to get the path cost f(n)=g(n)+h(n), and then put the path cost into the open container open_set;
[0088] S32, judging whether the open container open_set is empty;
[0089] If the open container open_set is empty, the path planning fails, and the process returns to step S31 to select a new adjacent node and recalculate the path cost f(n);
[0090] If the open container open_set is not empty, take the node with the smallest path cost f(n) from the open container open_set as the current node, and then add the current node to the closed container close_list;
[0091] S33, determine whether the current node is the end point of A*;
[0092] If the current node is the end point of A*, then output the local feasible path (i.e., a traceback information from the start point of A* to the end point of A* of the open container open_set), and then execute step S4; if the current node is not the end point of A*, then execute step S34;
[0093] S34. Calculate the path cost from the current node to one of the adjacent nodes:
[0094] If the adjacent node collides with the map after collision detection, it is discarded and step S36 is executed; if not, it is determined whether the adjacent node is in the closed container close_list;
[0095] If the adjacent node is not in the closed container close_list, calculate the path cost f(n) from the current node to the adjacent node, and then put the path cost f(n) corresponding to the adjacent node into the open container open_set, and then execute step S35;
[0096] If the adjacent node is in the closed container close_list, directly execute step S36;
[0097] S35, determining whether the adjacent node in step S34 is in the open container open_set;
[0098] If not, compare the path cost from the A* starting point to the adjacent node with the minimum value of the path cost in the current open container open_set; if the path cost from the A* starting point to the adjacent node is smaller, update the open container open_set, take the latest path cost corresponding to the adjacent node as the minimum value of the path cost in the open container open_set, and correspondingly add the adjacent node as the new current node to the closed container close_list, and then execute step S36; if the path cost does not change, execute step S36;
[0099] If yes, then put the adjacent node and its path cost into the open container open_set, and then execute step S36;
[0100] S36, determining whether all adjacent nodes of the current node in step S34 have been traversed;
[0101] If all adjacent nodes of the current node have been traversed, return to step S32;
[0102] If all adjacent nodes of the current node have not been traversed, return to step S34;
[0103] S4, linearly interpolating the two paths obtained in step S2 according to the resolution of the rasterized map of 0.025 meters, and then splicing them with the local feasible path obtained by the A* algorithm in step S33 to form a map path from the starting point to the target point;
[0104] S5, compare the path cost of the map path after splicing in step S4, or the path cost of the map path obtained in step S27 with the path cost of the map path in step S13, and update if the path cost is reduced, and finally form the following Figure 4 The planning effects shown; Figure 5 As shown, by comparison, it can be seen that the time consumption of the current planning is less than that of the previous round of planning. This method can adapt according to the number of iterations, enter a slow trial state or a divergent avoidance state and a fast recovery state, thereby improving the efficiency of path planning.
[0105] like Figure 6 The dynamic obstacle avoidance experiment was conducted using the settings shown in the figure, where the width of the active obstacle was set to twice the width of the mobile robot, the width of the channel was set to three times the width of the mobile robot, and the starting point and the end point were located at the lower left corner and the upper right corner of the venue, respectively. Figure 7 (a)- Figure 7 As shown in (f), in this dynamic obstacle avoidance experiment, the moving speed of the active obstacle is set to be 80% of the robot's moving speed. Figure 7 (a)- Figure 7 In (f), the time elapsed by the robot is set to 0 seconds, 4.2 seconds, 8.4 seconds, 12.6 seconds, 16.8 seconds, and 22 seconds, respectively. Figure 7 (a)- Figure 7 (f) It can be seen that the path planning method of this embodiment that integrates A* and improved RRT-Connect takes a much shorter time to give a complete path under the premise of dynamic obstacle avoidance than the time taken by using the A* algorithm alone. Obviously, the dynamic obstacle avoidance capability of the path planning method of this embodiment that integrates A* and improved RRT-Connect is better than that of using the A* algorithm alone.
[0106] Compared with the prior art, the path planning method of this embodiment that integrates A* and improved RRT-Connect has the following beneficial effects:
[0107] By adopting heuristic ideas to optimize the random sampling process based on the RRT-Connect algorithm, the improved RRT-Connect can quickly plan feasible paths in a large space. Combined with the advantage of the A* algorithm in accurately planning the optimal path under local complex structures, the algorithm is switched to adapt to different scenarios by using an adaptive adjustment method for the number of iterations, effectively improving the search speed and shortening the planning time, avoiding the problems of low search efficiency and long search time when using a search-based path planning method alone, avoiding continuous path failure, and blindly growing and expanding path costs when using a sampling-based path planning method alone; the improved RRT-Connect uses the randomness of the heuristic setting of sampling points to immediately re-search the path after the path planning fails due to changes in obstacles in the map, thereby realizing the function of dynamic obstacle avoidance.
[0108] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.
Claims
1. A path planning method integrating A* and improved RRT-Connect, characterized in that: The steps include: Initialization: load the map, set the number of iterations, iteration time, starting point, and target point; Planning a path according to the improved RRT-Connect algorithm: using the improved RRT-Connect algorithm to perform random sampling by setting heuristic sampling points in the map, performing spanning tree operations on the starting point tree and the target point tree to expand the corresponding nearest points, and then backtracking from the latest nearest points to the root nodes of the starting point tree and the target point tree, respectively, and judging whether the starting point tree and the target point tree are merged within a set number of iterations or iteration time; if so, directly splicing the backtracked path to obtain a map path from the starting point to the target point; if not, obtaining two paths, and then performing path search according to the A* algorithm; the heuristic sampling points are set to use the starting point tree and the target point tree as sampling areas with a set probability, and use the map area outside the starting point tree and the target point tree as sampling areas with a set probability; Path search based on A* algorithm: Use A* algorithm to search for the local feasible path with the minimum path cost from the starting point tree to the target point tree on the nearest points corresponding to the starting point tree and the target point tree; Three-segment path splicing: Splice the two segments of the path obtained by planning the path according to the improved RRT-Connect algorithm with the local feasible path obtained by path search according to the A* algorithm to obtain the map path from the starting point to the target point; The steps for planning a path according to the improved RRT-Connect algorithm are as follows: Set the starting point and expansion step size for the starting point tree and the target point tree respectively, and then use the heuristic sampling point setting to randomly generate sampling points on the map; Searching for the nearest neighbor points of the starting point tree and the target point tree respectively, wherein the nearest neighbor points are the sampling points of the starting point tree or the target point tree that are closest to each other; Starting from the nearest point, set the starting point tree and the target point tree to grow temporary nodes in the direction of the sampling point with an extended step length, and establish a temporary connection between the nearest point and the temporary node; Whether the temporarily connected line falls into the range of obstacles in the map or exceeds the map boundary; If yes, disconnect the temporary connection; if no, add the temporary node to the starting point tree or the target point tree accordingly; Search for the optimal parent node of the temporary node to update the starting point tree or the target point tree accordingly; Determine whether the starting point tree and the target point tree are merged to output the map path.
2. The path planning method integrating A* and improved RRT-Connect according to claim 1, characterized in that: The initialization also includes searching for a path from a starting point to a target point using an A* algorithm when performing path planning for the first time, and using the obtained map path as a reference for a map path obtained by splicing the three path segments.
3. The path planning method integrating A* and improved RRT-Connect according to claim 1, characterized in that: The step of searching for the optimal parent node of the temporary node to update the starting point tree or the target point tree accordingly specifically includes: Calculate the sampling point with the smallest path cost to the temporary node and select it as the candidate for the optimal parent node; establish the connection between the candidate optimal parent node and the temporary node and determine whether the connection falls into the obstacle range in the map or exceeds the map boundary; If yes, disconnect the connection between the candidate optimal parent node and the temporary node; If not, it is taken as the optimal parent node and merged into the starting point tree or the target point tree accordingly.
4. The path planning method integrating A* and improved RRT-Connect according to claim 1, characterized in that: The determining whether the starting point tree and the target point tree merge specifically includes: If they converge, the latest temporary nodes of the starting point tree and the target point tree are used to trace back to the starting point and the target point respectively, and then the two traced paths are spliced to obtain the map path from the starting point to the target point; If there is no convergence, determine whether the number of iterations or iteration time exceeds the set value during initialization; If exceeded, restart the sampling point setting; If not, then the step of performing path search according to the A* algorithm is performed.
5. The path planning method integrating A* and improved RRT-Connect according to claim 1, characterized in that: The steps of performing path search according to the A* algorithm are as follows: Mark the latest nearest point in the starting point tree as the A* starting point, and the latest nearest point in the target point tree as the A* end point, then calculate the path cost of all paths from the A* starting point to the A* end point; take the path with the smallest path cost as the local feasible path from the A* starting point to the A* end point.
6. The path planning method integrating A* and improved RRT-Connect according to claim 5, characterized in that: The calculation of the path cost of all paths from the A* starting point to the A* end point specifically includes: Initialize the open and closed containers in the A* algorithm; Mark the cumulative cost of the starting point of A* as zero, calculate the estimated cost between the two nodes, add the cumulative cost and the estimated cost to get the path cost, and then put the path cost into the open container; Determine whether the open container is empty; if the open container is empty, the path planning fails, and the path cost is recalculated by selecting a new adjacent node; if the open container is not empty, the node with the smallest path cost is taken from the open container as the current node, and then the current node is added to the closed container; Determine whether the current node is the A* end point; if the current node is the A* end point, output a local feasible path, and then perform the three-segment path splicing; if the current node is not the A* end point, traverse the adjacent nodes of the current node to continue the path search.
7. The path planning method integrating A* and improved RRT-Connect according to claim 6, characterized in that: The traversing of the adjacent nodes of the current node to continue the path search specifically includes: Calculate the path cost from the current node to one of the adjacent nodes; Determine whether the adjacent node is in the open container; if not, compare the path cost from the A* starting point to the adjacent node with the minimum path cost in the current open container; if the path cost from the A* starting point to the adjacent node is smaller, update the open container, take the latest path cost corresponding to the adjacent node as the minimum path cost in the open container, and add the adjacent node as the new current node to the closed container; if it is, put the adjacent node and its path cost into the open container; Determine whether all adjacent nodes of the current node have been traversed; if all adjacent nodes of the current node have been traversed, return to the step of determining whether the open container is empty; if all adjacent nodes of the current node have not been traversed, return to the step of calculating the path cost from the current node to its next adjacent node.
