Mechanical arm motion planner path optimization obstacle avoidance method based on improved Inform-RRT*

Abstract

The invention relates to the field of mechanical arm path planning, in particular to a path optimization obstacle avoidance method based on an improved Inform-RRT * mechanical arm motion planner. Firstly, parameter information required by path planning is acquired, then an initial step length is determined according to an environment, a random sampling point is generated, a nearest node is found, and a new node is generated according to the step length; continuously updating path points and father nodes, and optimizing path cost; by using the rewiring process in the RRT * algorithm, the radius search of the fixed step length is changed into the dynamic step length, the path length is shortened, and the algorithm efficiency is improved. In combination with an APF function, an artificial potential field function in the APF function is fused into an Inform-RRT * algorithm, and a good convergence effect of the algorithm is used for making up the defects that the number of nodes is too large and convergence is slow before the traditional Inform-RRT * forms an elliptical convergence domain. And finally, the improved path planning obstacle avoidance algorithm is added into an ompl planning library, and a path planning obstacle avoidance experiment is completed in ROS MoveIt in combination with a UR3 mechanical arm model. The number of sampling points is effectively reduced, and the working efficiency of the mechanical arm is improved.

Description

Technical Field

[0001] This invention relates to the field of robotic arm path planning, and more particularly to a method based on an improved Informed-RRT. ＊ Obstacle avoidance method for path optimization in robotic arm motion planner. Background Technology

[0002] Robots can be divided into mobile robots and articulated robots (also known as robotic arms). Generally, the working environment of a robotic arm is not a completely free and unconstrained space, but rather a constrained space filled with various obstacles. With the diversification and complexity of intelligent manufacturing scenarios (such as industrial, agricultural, and military fields), manual teaching methods are clearly inefficient; therefore, researching efficient path planning methods is essential. Traditional path planning methods are slow and have many shortcomings, while a good path planning algorithm can enable the robotic arm to complete tasks with the shortest path and minimum energy consumption, thereby improving work efficiency and quality, reducing joint wear, reducing operating costs, and preventing collisions between the robotic arm and obstacles or other equipment.

[0003] Although many scholars have proposed numerous algorithmic theories in the field of robot path planning, such as the visual diagram method and the grid method, these algorithms have improved the efficiency of path planning to some extent. However, obstacle avoidance path planning algorithms for robotic arms still have many shortcomings. Especially in high-dimensional spaces, the path planning technology for robotic arms is not yet mature. Therefore, from both academic research and industrial application perspectives, research on obstacle avoidance path planning algorithms for robotic arms is of great significance for promoting the intelligent development of robotic arms and improving the level of social productivity.

[0004] Because RTT (Rapidly Exploring Random Tree) is more suitable for high-dimensional spaces, it is a sampling-based random sampling algorithm for robot path planning. The basic idea is to use the starting point as the root node and grow a search tree in space through continuous random sampling and collision detection. Each node has a unique parent node. When a leaf node expands to the vicinity of the target point, a collision-free path from the starting point to the target point can be obtained by continuously backtracking the parent node. However, the sampling in the RRT algorithm is simple random uniform sampling. For environments with many obstacles and narrow gaps, RRT's exploration ability is poor, and the RRT algorithm does not possess optimality. LaValle implemented a simple bias during sampling, with a certain probability of sampling near the target point in each sampling, guiding the algorithm tree to grow towards the target point. To address the problem of the RRT algorithm's solution not possessing optimality, Sertac and Emilio proposed RRT... ＊ This refers to the RRT algorithm, which possesses asymptotic optimality. RRT ＊The algorithm adds a pruning module to the RRT, introducing a cost function, reselecting parent nodes, and rewiring a random tree. When a new node is about to be added, instead of directly connecting it to the nearest node in the tree, it searches for the optimal parent node within a certain range and then reconnects nodes within that range that meet certain rules. However, the fixed radius pruning module's effectiveness is significantly reduced in the early stages of the search due to the small number of sampled nodes; and in the later stages, when the number of nodes is large, multiple path cost calculations are performed, resulting in wasted computational power and time.

[0005] Traditional Informed-RRT ＊ Simply restricting the sampling area does not completely solve the problem of unordered sampling. Furthermore, the algorithm is still completely unordered in its search until it finds the first feasible solution, which leads to the generation of many redundant paths and reduces the convergence speed of the algorithm for the optimal path. Summary of the Invention

[0006] To address the problems of disordered sampling and redundant paths in existing technologies, this invention provides a method based on improved Informed-RRT. ＊ Obstacle avoidance method for path optimization in robotic arm motion planner.

[0007] According to the first aspect of this disclosure, a method based on an improved Informed-RRT is provided. ＊ The path optimization and obstacle avoidance methods for robotic arm motion planners include: S1: Establish a six-degree-of-freedom robotic arm model, and use a virtual machine and MATLAB to initialize a random tree, initial node Xstart, step size d, and target point Xgoal; S2: Based on the gravitational potential field and the repulsive potential field, the first sampling point Xrand1 is generated by the random tree; S3: Determine whether the end effector of the robotic arm collides with an obstacle. If yes, proceed to S2 for resampling; otherwise, proceed to S4. S4: Based on Xrand1, obtain the first new node Xnew1; S5: Using the dynamic pruning module, backtrack the path at Xnew1 and at the sampling points within the dynamic pruning domain with Xnew1 as the center and R as the radius. Determine whether the path cost from Xstart to Xnew1 is smaller than the path cost from Xstart to the sampling point. If so, update Xnew1; otherwise, do not change Xnew1. S6: Using Xnew1 as the parent node, backtrack the path within the dynamic pruning domain; S7: Determine whether the path cost connecting the nodes in Xstart, Xnew1, and the dynamic pruning domain in sequence is smaller than the path cost from Xstart to Xnew1. If so, rewire with the minimum path cost; otherwise, do not change the path. S8: Determine whether the distance from the end point of the path obtained in S7 to Xgoal is less than the threshold. If so, connect the end point of the path and Xgoal to generate an elliptical sampling threshold with the line connecting the end point of the path and Xgoal as the major axis, and proceed to S9. If not, resample in S2. S9: Sample within the elliptical sampling threshold to obtain the second sampling point Xrand2; S10: Based on Xrand2, obtain the second new node Xnew2, determine whether the end of the robotic arm collides with an obstacle. If so, resample in S9; otherwise, use the same method as S5-S7 to backtrack the path. S11: Determine if the maximum number of iterations has been reached. If yes, end the iteration and obtain the optimal obstacle avoidance path. If no, return to S9 and continue the operation until the optimal obstacle avoidance path is obtained.

[0008] According to an embodiment of this disclosure, the first sampling point Xrand1 is determined by the gravitational potential field function generated by the target point Xgoal and the repulsive potential field function generated by the obstacle, and is generated randomly. When the distance between the end effector of the robotic arm and the obstacle exceeds the range of the repulsive potential field, Xrand1 is not affected by the repulsive potential field.

[0009] According to embodiments of this disclosure, the above-mentioned gravitational potential field function is as shown in formula (1): (1) in, It is a gravitational potential field. It is the gravitational potential field gain. It is the distance between the sampling point and the target point Xgoal.

[0010] According to an embodiment of this disclosure, the above-mentioned repulsive potential field function is as shown in formula (2): (2) in, It is a repulsive potential field. It is the gain of the repulsive potential field. It is the distance between the sampling point and the obstacle. It is the range of the repulsive potential field of the obstacle.

[0011] According to an embodiment of this disclosure, the process of obtaining the first new node Xnew1 in S4 above includes the following steps: S4.1: Traverse all nodes in the current random tree, find the first node Xnear1 that is closest to Xrand1, and update Xnear1 by calculating the distance from each node in the tree to Xrand1 and selecting the node with the smallest distance. S4.2: Point Xnear1 to Xrand1 to obtain a vector. Then, normalize the vector and multiply it by the step size d to obtain a new vector of length d. Use the endpoint of the new vector as the first new node Xnew1. In S10, the method for obtaining the first new node Xnew1 is used to obtain the second new node Xnew2.

[0012] According to the embodiments of this disclosure, in the above S6, the backtracking path is implemented by searching for other sampling points in a dynamic pruning domain with Xnew1 as the center and R as the radius, and sequentially connecting the sampling points in the domain with Xnew1 to obtain the backtracking path; the radius R is calculated as shown in formula (3): (3) Where k is the step size adjustment coefficient, ceil() is MATLAB's built-in function for rounding up, and numel(nodes_list) defines the total number of nodes in the current random tree.

[0013] According to an embodiment of this disclosure, in S7, the node that generates the minimum path cost is updated as the latest parent node of Xnew1, and then Xnew1 is used as the parent node to perform rerouting with the minimum path cost.

[0014] A second aspect of this disclosure provides a computer apparatus including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0015] A third aspect of this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0016] A fourth aspect of this disclosure provides a computer program product including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0017] According to embodiments of this disclosure, a method based on an improved Informationed-RRT is provided. ＊ The path optimization and obstacle avoidance method of robotic arm motion planner, in addition to the traditional Informed-RRT ＊In the algorithm, using a fixed-range pruning module to reconnect parent nodes and rewire effectively improves path quality, but it doesn't significantly reduce the search time. In the early stages of random tree generation, the sampling point density is too low, making the pruning module less meaningful, as there may be very few, or even none, sampling points within a pruning region of radius R centered on Xnew. However, in the later stages of random tree generation, as the sampling point density increases, the same region size significantly increases the computational load and memory consumption. Therefore, using a dynamic pruning module is essential, as it can optimize the path while also shortening the search time. In traditional Informed-RRT... ＊ While the elliptic convergence region algorithm restricts the sampling points to an elliptic region during path searching, enabling rapid convergence to the optimal path, the search remains unordered until the first path solution is found, potentially leading to path redundancy and wasting computational resources. Therefore, the APF function is introduced to quickly find the first path solution. The APF algorithm constructs an attractive potential field for the target point and a repulsive potential field for obstacles. The attractive potential field generates an attractive force, while the repulsive potential field generates a repulsive force; these two forces work together to guide the robot to the target point. The attractive potential field is generated by the target point, and the repulsive potential field is constructed similarly, generated by obstacles. When the distance to an obstacle exceeds the range of the repulsive potential field, the current sampling point is unaffected by it. This invention improves the Informed-RRT. ＊ The algorithm can achieve rapid convergence using an artificial potential function to find the initial path solution even when no initial solution is found. Furthermore, the addition of a dynamic pruning module changes the traditional Informed-RRT approach. ＊ The algorithm addresses the problem of wasted computing power, significantly improves convergence speed, reduces redundant paths, and enables precise grasping of obstacle avoidance path planning for a six-degree-of-freedom robotic arm. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is an embodiment of the present invention based on an improved Informed-RRT. ＊ Flowchart of the path optimization and obstacle avoidance method for robotic arm motion planner; Figure 2 This is a schematic diagram of the UR3 six-axis robotic arm model built in SOLIDWORKS 2019 in this embodiment of the invention, with each reference axis configured. Figure 3 This is a schematic diagram of the model presented in rviz after importing the URDF file into the ROS system and running the launch file in the RDF file in an embodiment of the present invention. Figure 4 This is the RRT in the embodiment of the present invention. ＊ A schematic diagram illustrating the path planning and obstacle avoidance performance of the algorithm after 5000 samplings; Figure 5 In this embodiment of the invention, the APF function is added to RRT. ＊ A schematic diagram illustrating the path planning and obstacle avoidance effect; Figure 6 This is an embodiment of the Informed-rrt invention. ＊ A schematic diagram illustrating the boundary definition of the elliptic guided domain in the algorithm; Figure 7 This is the Informed-RRT embodiment of the present invention. ＊ A schematic diagram illustrating the path planning and obstacle avoidance performance of the improved algorithm after 349 samplings; Figure 8 This is an example of Informed-RRT under different environments in the embodiments of the present invention. ＊ A schematic diagram illustrating the path planning and obstacle avoidance performance of the improved algorithm after 172 samplings; Figure 9 This is a schematic diagram illustrating the effect of the algorithm on path planning and obstacle avoidance after adding the APF function in a three-dimensional environment in this embodiment of the invention. Figure 10 This is a simulation experiment diagram of the six-axis robotic arm in an embodiment of the present invention after incorporating an improved path planning and obstacle avoidance algorithm. Detailed Implementation

[0019] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] Example 1 Please refer to Figure 1 , Figure 1 This is an embodiment of the present invention based on an improved Informed-RRT. ＊ The flowchart illustrates the path optimization and obstacle avoidance method for the robotic arm motion planner. The hardware environment for implementing this method is as follows: a DELL G3 laptop equipped with an Intel i7-9750H CPU is used as the program running and simulation device. The software environment for implementing this method is as follows: an Ubuntu 18.04 system with a modified firmware and a recompiled kernel version of 4.4.0 is used, and Moveit! Setup Assistant is installed on the Ubuntu system.

[0021] This invention first obtains the parameter information required for path planning, then determines the initial step size based on the environment, generates random sampling points, finds the nearest node, and generates new nodes according to the step size; it continuously updates path points and parent nodes to optimize path cost; and utilizes RRT (Real-Time Tracking). ＊ The rewiring process in the algorithm changes the fixed-step radius search to a dynamic step size, shortening the path length and improving algorithm efficiency. Combined with the APF function, the artificial potential field function is integrated into Informed-RRT. ＊ In the algorithm, its excellent convergence performance is used to compensate for the shortcomings of the traditional Informed-RRT. ＊ Before forming the elliptical convergence region, the algorithm suffers from the drawback of having too many nodes and slow convergence. Finally, the improved path planning and obstacle avoidance algorithm was added to the ompl planning library and, combined with the UR3 robotic arm model, path planning and obstacle avoidance experiments were conducted in ROSMoveIt.

[0022] The method specifically includes the following steps: S1: As Figure 2 As shown, the .SLDASM file of the UR3 six-DOF robotic arm is imported into SOLIDWORKS 2019 to configure the reference axes and coordinate system. After configuration, it is converted into a URDF file for the robotic arm, in which the appearance, collision properties, and inertial properties of the robotic arm are defined. This is to facilitate subsequent control experiments for robotic arm simulation in ROS Moveit! Setup Assistant.

[0023] S2: Open the virtual machine, enter the Ubuntu 18.04 system, use the Moveit! Setup Assistant tool to configure the URDF file obtained by S1, and open and view the configured six-DOF robotic arm in Rviz, such as... Figure 3 As shown in the figure, joint1, joint2, joint3, joint4, joint5, and joint6 represent the six joints of the robotic arm, Randomize represents the radius, and Center represents the center. This is for experimental simulation verification using Moveit! under the ROS operating system in S12.

[0024] S3: Return to the Windows system and open MATLAB 2022 to write the path planning algorithm program. Using MATLAB, initialize the random tree, the initial node Xstart, the step size d, and the target point Xgoal. In this embodiment, the step size d can be 30. The starting node and target point can be freely set according to actual needs. Load environmental obstacles into MATLAB to demonstrate the obstacle avoidance effect of the path search algorithm.

[0025] S4: Based on the gravitational and repulsive potential fields of the sampling function APF, the random tree generates a random sampling point Xrand1.

[0026] S5: Based on the sampling point Xrand1, obtain the new node Xnew1.

[0027] The process of obtaining Xnew1 is as follows: S5.1: Traverse all nodes of the current random tree, find the node Xnear1 that is closest to the random sampling point Xrand1, and update Xnear by calculating the distance from each node in the tree to Xrand1 and selecting the node with the smallest distance.

[0028] S5.2: Normalize and expand to obtain a new node Xnew1. That is, use Xnear1 to point to a new vector in the direction of Xrand1. Then, normalize the vector and multiply it by d, that is, scale this unit vector using a step size d, to obtain a vector of length d. This is used to obtain the new node Xnew1.

[0029] S6: Using the dynamic pruning module, backtrack the path at the new node Xnew1 and at the sampling points within the dynamic pruning domain with Xnew1 as the center and R as the radius; The implementation process of the backtracking path is as follows: Within the dynamic pruning region centered at Xnew1 and with radius R, other sampling points are searched. By sequentially connecting the sampling points within the region to Xnew1, the backtracking path can be obtained. The radius R is calculated as shown in formula (1): (1) Where k is the step size adjustment coefficient, which is obtained empirically, and in this embodiment, k is set to 30. ceil() is a built-in MATLAB function for rounding up, and numel(nodes_list) defines the total number of nodes in the current random tree.

[0030] Determine if the path cost from Xstart to Xnew1 is smaller than the path cost from Xstart to the sampling point. If so, update Xnew1; otherwise, do not change Xnew1. The actual path length generated from the starting node to the target point represents the path cost.

[0031] S7: Using Xnew1 as the parent node, backtrack the path within the dynamic pruning domain.

[0032] S8: Determine whether the path cost connecting the nodes in Xstart, Xnew1, and the dynamic pruning domain in sequence is smaller than the path cost from Xstart to Xnew1. If so, rewire with the minimum path cost; otherwise, do not change the path. S9: Determine whether the distance from the path endpoint obtained in S8 to Xgoal is less than the threshold. If yes, connect the path endpoint and Xgoal to generate an elliptical sampling threshold with the line connecting the path endpoint and Xgoal as the major axis, and proceed to S9. If no, proceed to S2 for resampling.

[0033] S10: Sampling is performed within the elliptical sampling threshold to obtain sampling point Xrand2; S11: Based on Xrand2, obtain the second new node Xnew2, determine whether the end of the robotic arm collides with an obstacle. If yes, resample in S10; otherwise, use the same method as S6-S8 to backtrack the path. S11: Determine if the maximum number of iterations has been reached. If yes, end the iteration and obtain the optimal obstacle avoidance path. If no, return to S9 and continue the corresponding operation until the optimal obstacle avoidance path is obtained.

[0034] In traditional Informationd-RRT ＊ In the algorithm, the generated sampling points are disordered and random, such as... Figure 4 As shown, the green path represents the currently found optimal path from the starting node to the target node, the blue line represents the tree structure generated by the algorithm expansion (i.e., the line connecting parent and child nodes), the red line represents the newly connected line after rewiring, and the black part represents obstacles. The red points are sampling points. Figure 4 As can be seen, the sampling points are randomly located, which wastes actual computation time and consumes a lot of computing power. Therefore, this invention introduces a sampling method that combines the APF function and the elliptic convergence region to improve the traditional Informed-RRT. ＊ The algorithm was improved by adjusting the sampling points. of The generation process is divided into two stages: In the first stage, before the path reaches the target point Xgoal, the APF function is used for sampling. The sampling point Xrand1 is jointly determined by the gravitational potential field function generated by the target point Xgoal and the repulsive potential field function generated by the obstacle; The gravitational potential field function is shown in equation (2): (2) in, It is a gravitational potential field. It is the gravitational potential field gain. It is the distance between the current sampling point and the target point Xgoal.

[0035] The repulsive potential field function is shown in formula (3): (3) in, It is a repulsive potential field. It is the gain of the repulsive potential field. It is the distance between the current sampling point and the obstacle. It is the range of the repulsive potential field of the obstacle.

[0036] When the distance between the end effector of the robotic arm and the obstacle exceeds the range of the repulsive potential field, the current sampling point is not affected by the repulsive potential field.

[0037] like Figure 5 As shown, the introduction of the artificial potential field function (APF) enables the path to quickly converge towards the target point before reaching it, ensuring that the generated path is near the optimal path. This reduces the initial path generation time and improves the efficiency of Informed-RRT. ＊ The computation speed of the algorithm. Figure 5 In the diagram, the red part represents the path from the starting point (0,0) to the ending point (1000,1000), the dark green square part represents obstacles, and the blue part represents the line from the newly generated node to its parent node.

[0038] After the initial path is generated, sampling enters the second stage, which involves restricting the sampling points to an elliptical domain. The definition of the elliptical domain is as follows: Figure 6 As shown, Xstart is the selected initial node (i.e., the starting point), Xgoal is the selected target point (i.e., the ending point), and c min c is the straight-line distance from the starting point to the ending point, which is also the focal length of the ellipse. best The major axis of the ellipse is determined by the path cost, which is the length of the initial solution found by the path. Indicates the minor axis of the ellipse. Improved Informed-RRT ＊ The algorithm generates sampling points around the path in an ordered manner.

[0039] The improved Informationd-RRT ＊ The algorithm runs in different environments, randomly placing various randomly shaped obstacles, and obtains results such as... Figure 7 , Figure 8 The actual effect shown. Figure 7 , Figure 8 In the diagram, black represents obstacles, red represents newly connected lines after rewiring, blue represents the tree structure generated by the algorithm (connecting parent and child nodes), and green represents the currently found optimal path from the starting point to the target point. This is to verify the improved Informed-RRT. ＊ The algorithm's planning capability in high-dimensional space was demonstrated using MATLAB as the simulation platform. A 900×900×900 three-dimensional space was created as the sample space in the data space, and multiple obstacles were randomly generated. The results are as follows: Figure 9As shown, the red line is the optimal path from the starting node to the target point generated in the current iteration, the blue line is the connection from Xnew to its parent node, the green line is the rewire line, and the red cube, green cylinder, and blue sphere represent obstacles.

[0040] S12: Reopen Ubuntu 18.04 and choose to reinstall Moveit! from source, including the improved Informed-RRT. ＊ The algorithm has been added to the ompl path planning algorithm library. For a robotic arm to travel from its starting point to its destination, it must avoid obstacles. In the Rviz simulation environment, obstacles and initial / final states are as follows: Figure 10 As shown, the yellow robotic arm is in the target pose, the light-colored robotic arm is in the initial pose, and the green represents the obstacles. This completes the simulation verification of the robotic arm path planning in ROS.

[0041] Example 2 A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0042] Example 3 A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above method.

[0043] Example 4 A computer program product includes a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0044] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method based on improved Informed-RRT ＊ The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, Includes the following steps: S1: Using ROS, establish a URDF model of a six-degree-of-freedom robotic arm, and then use MATLAB to initialize the random tree, initial node Xstart, step size d, and target point Xgoal; S2: Based on the gravitational potential field and the repulsive potential field, the first sampling point Xrand1 is generated by the random tree; S3: Determine whether the end effector of the robotic arm collides with an obstacle. If yes, proceed to S2 for resampling; otherwise, proceed to S4. S4: Based on Xrand1, obtain the first new node Xnew1; S5: Using the dynamic pruning module, backtrack the path at Xnew1 and at the sampling points within the dynamic pruning domain with Xnew1 as the center and R as the radius. Determine whether the path cost from Xstart to Xnew1 is smaller than the path cost from Xstart to the sampling point. If so, update Xnew1; otherwise, do not change Xnew1. S6: Using Xnew1 as the parent node, backtrack the path within the dynamic pruning domain; S7: Determine whether the path cost connecting the nodes in Xstart, Xnew1, and the dynamic pruning domain in sequence is smaller than the path cost from Xstart to Xnew1. If so, rewire with the minimum path cost; otherwise, do not change the path. S8: Determine whether the distance from the end point of the path obtained in S7 to Xgoal is less than the threshold. If so, connect the end point of the path and Xgoal to generate an elliptical sampling threshold with the line connecting the end point of the path and Xgoal as the major axis, and proceed to S9. If not, resample in S2. S9: Sample within the elliptical sampling threshold to obtain the second sampling point Xrand2; S10: Based on Xrand2, obtain the second new node Xnew2, determine whether the end of the robotic arm collides with an obstacle. If so, resample in S9; otherwise, use the same method as S5-S7 to backtrack the path. S11: Determine if the maximum number of iterations has been reached. If yes, end the iteration and obtain the optimal obstacle avoidance path. If no, return to S9 and continue the operation until the optimal obstacle avoidance path is obtained.

2. A method based on improved Informed-RRT according to claim 1 ＊ The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, In S2, the first sampling point Xrand1 is determined by the gravitational potential field function generated by the target point Xgoal and the repulsive potential field function generated by the obstacle, and is generated randomly. When the distance between the end effector of the robotic arm and the obstacle exceeds the range of the repulsive potential field, Xrand1 is not affected by the repulsive potential field.

3. A method based on the improved Information- according to claim 2 The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, The gravitational potential field function is shown in equation (1): （1） in, It is a gravitational potential field. It is the gravitational potential field gain. It is the distance between the sampling point and the target point Xgoal.

4. A method based on improved Informed-RRT according to claim 2 ＊ The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, The repulsive potential field function is shown in equation (2): （2） in, It is a repulsive potential field. It is the gain of the repulsive potential field. It is the distance between the sampling point and the obstacle. It is the range of the repulsive potential field of the obstacle.

5. A method based on improved Informed-RRT according to claim 1 ＊ The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, In S4, the process of obtaining the first new node Xnew1 includes the following steps: S4.1: Traverse all nodes in the current random tree, find the first node Xnear1 that is closest to Xrand1, and update Xnear1 by calculating the distance from each node in the tree to Xrand1 and selecting the node with the smallest distance. S4.2: Point Xnear1 to Xrand1 to obtain a vector. Then, normalize the vector and multiply it by the step size d to obtain a new vector of length d. Use the endpoint of the new vector as the first new node Xnew1. In S10, the method for obtaining the first new node Xnew1 is used to obtain the second new node Xnew2.

6. A method based on improved Informed-RRT as described in claim 5 ＊ The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, In S6, the backtracking path is implemented by searching for other sampling points within a dynamic pruning domain centered at Xnew1 and with radius R, and then connecting the sampling points within the domain to Xnew1 to obtain the backtracking path; the radius R is calculated as shown in formula (3): （3） Where k is the step size adjustment coefficient, ceil() is MATLAB's built-in function for rounding up, and numel(nodes_list) defines the total number of nodes in the current random tree.

7. A method based on improved Informed-RRT according to claim 5 ＊ The path optimization and obstacle avoidance method for robotic arm motion planners is characterized by, In S7, the node that generates the minimum path cost is updated as the latest parent node of Xnew1, and then Xnew1 is used as the parent node to rewire the system with the minimum path cost.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes a computer program to implement the improved Informed-RRT based on any one of claims 1-7. ＊ The steps of the path optimization and obstacle avoidance method for robotic arm motion planners.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the improved Informed-RRT based on any one of claims 1-7. ＊ The steps of the path optimization and obstacle avoidance method for robotic arm motion planners.

10. A computer program product, characterized in that, Includes a computer program or instructions that, when executed by a processor, implement the improved Informed-RRT based on any one of claims 1-7. ＊ The steps of the path optimization and obstacle avoidance method for robotic arm motion planners.

Images