Industrial robot grabbing obstacle avoidance path planning method based on RRT-TAGB algorithm

Abstract

The invention discloses an industrial robot grabbing obstacle avoidance path planning method based on an RRT-TAGB algorithm, and belongs to the field of mechanical assembly. Firstly, a self-adaptive target bias sampling strategy is provided for the problem that a traditional RRT algorithm is large in search randomness, the target sampling probability is determined through the volume relation between an obstacle and a grabbing work space, a random tree is made to be expanded to a target point more quickly, and the guidance quality of the algorithm in the search process is improved; secondly, in order to effectively solve the problem that path planning fails due to the fact that local optimum is prone to occurring in the dense obstacle environment, an artificial potential field method is used for conducting local path planning, and the positions of nodes are adjusted again to obtain an optimized path; then, aiming at the problem that invalid nodes on the path are too many, a greedy strategy is adopted to remove redundant nodes of the path, and the length of the path is reduced; and finally, aiming at the problem of poor path quality, performing smoothing operation on the path by using a cubic B-spline curve, and removing unnecessary corners so as to improve the quality of the path. According to the RRT-TAGB algorithm adopted by the invention, the guidance quality of random tree growth is improved, redundant nodes on a path are greatly reduced, the search time is shortened, the final path is smoother, the overall obstacle avoidance path quality is higher, and the adaptive capacity of grabbing and obstacle avoidance of the industrial robot in a complex environment is ensured.

Description

Technical Field

[0001] The present invention relates to the field of mechanical assembly, and in particular to an industrial robot grasping obstacle avoidance path planning method based on RRT-TAGB algorithm. Background Art

[0002] In industrial production lines, a very common scenario is the grasping and placement of parts by industrial robots. Therefore, it is crucial to plan a spatial path from the starting point to the target point for the end effector of the industrial robot in the grasping operation space to ensure that the end effector of the robot can avoid all obstacles in the operation space during the grasping operation. However, the grasping environment is often characterized by high dimensionality, complexity, and variability, which requires higher requirements for the path planning algorithm while ensuring safety. The quality of grasping obstacle avoidance path planning directly affects the working efficiency of the robot arm, and even affects the loss of the robot arm itself and its safety during work.

[0003] At present, researchers at home and abroad have developed a variety of obstacle avoidance path planning algorithms. Common obstacle avoidance path planning methods include Dijkstra algorithm, A* algorithm, Artificial Potential Field (APF), Probabilistic Road Maps (PRM) algorithm and Rapidly-exploring Random Trees (RRT).

[0004] However, before path planning, the three methods of Dijkstra algorithm, A* algorithm and artificial potential field method need to accurately model and describe the obstacles in the environment, and the computational complexity grows exponentially in complex environments, and they are prone to fall into local minima, with low path generation efficiency and poor path quality, which are not suitable for path planning in high dimensions. Although the PRM algorithm can be applied to high-dimensional space path planning, its performance depends to a large extent on the number and distribution of sampling points. Especially in complex environments, a large number of sampling points may be required to build an effective path map, which increases the computational complexity and time cost, and is not suitable for the application scenario of industrial robots quickly grasping and avoiding obstacles. The RRT algorithm does not need to preprocess the space when exploring unknown areas. By generating a tree structure through random sampling, RRT can quickly explore feasible paths and avoid obstacles. Therefore, it is widely used in the field of path planning in high-dimensional space, especially for grasping and avoiding obstacles path planning of industrial robots.

[0005] Although the RRT algorithm has good performance in high-dimensional space, it still has problems such as large search randomness, easy to fall into local optimality, many invalid nodes and poor path quality. Therefore, how to further optimize the RRT algorithm to obtain more efficient and accurate path planning in complex grasping and obstacle avoidance tasks has become an urgent problem to be solved. Summary of the invention

[0006] In view of the above problems, the present invention proposes a RRT-TAGB (RRT-Target-Artificial-Greedy-B-Spline curves) obstacle avoidance path planning algorithm to achieve a more efficient and accurate path in the grasping and obstacle avoidance task. First, in view of the problem of high randomness in the original algorithm search, an adaptive target bias sampling strategy is proposed. The target sampling probability is determined by the volume relationship between the obstacle and the grasping workspace, so that the random tree can be expanded to the target point faster, and the guidance of the algorithm in the search process is improved; secondly, in order to effectively solve the problem of being easily trapped in the local optimum in a dense obstacle environment, resulting in the failure of path planning, the artificial potential field method is used for local path planning, and the position of the node is readjusted to obtain the optimized path; then, in view of the problem of too many invalid nodes on the path, the greedy strategy is used to remove the redundant nodes of the path and reduce the length of the path; finally, in view of the problem of poor path quality, the cubic B-spline curve is used to smooth the path and remove unnecessary corners to improve the quality of the path and obtain a smooth obstacle avoidance path with higher quality. The RRT-TAGB algorithm adopted in the present invention improves the guidance of random tree growth, greatly reduces redundant nodes on the path, shortens the search time, makes the final path smoother, and improves the overall obstacle avoidance path quality, thereby ensuring the adaptability of the industrial robot to grasp and avoid obstacles in complex environments.

[0007] The technical solution adopted by the present invention is: a method for grasping obstacle avoidance path planning of an industrial robot based on the RRT-TAGB algorithm, which specifically includes the following steps:

[0008] Step 1: Use a depth camera to obtain the grasping workspace of the industrial robot;

[0009] Step 2: Simplify the obstacles in the workspace into spherical obstacles and rectangular obstacles. Create obstacle maps;

[0010] Step 3: Initialize variables, determine the starting point, target point and search step size;

[0011] Step 4: Determine the adaptive target bias probability value based on the obstacle information in the workspace. And perform adaptive target bias sampling;

[0012] Step 5: Iterate the random tree, select a closest node from the tree, and then generate a new node by sampling;

[0013] Step 6: Local path planning: adjust the position of the newly generated nodes by artificial potential field method to perform local path planning;

[0014] Step 7: Collision detection, check whether there is an obstacle between the new node and the nearest node in the tree. If there is no collision, add the new node to the tree; if there is a collision, return to step 5 and resample;

[0015] Step 8: Update the path. Add new nodes without collision to the tree. If the distance between the new node and the end point is smaller than the threshold, it is considered that the path has been found and the loop is exited.

[0016] Step 9: Remove redundant nodes. After the path is generated, use the greedy algorithm to optimize it, delete the redundant nodes on the path, and obtain the optimized path.

[0017] Step 10: Use the cubic B-spline curve to smooth the optimized path to obtain the final grasping and obstacle avoidance path of the industrial robot.

[0018] Preferably, the determination of the adaptive target bias probability value in step 4 is specifically calculated by the following formula: Where λ represents the obstacle volume correction coefficient, which is used to improve the error between the obstacle equivalent model volume and the actual volume, and its value is 0.8~1.2; V c Represents the volume of the rectangular obstacle; V q Represents the volume of the spherical obstacle; V k Represents the volume of the working space at the end of an industrial robot.

[0019] Preferably, in step 4, adaptive target bias sampling is performed, and the specific implementation process is as follows: According to the adaptive target bias probability P target (value range is 0~1), each time a node is expanded, a new probability value P is uniformly and randomly generated between 0 and 1; when P>P target When , the sampling point is still randomly selected in the workspace; otherwise, the target point X is directly selected. goal as the sampling point.

[0020] Preferably, the step 6 adjusts the position of the initially generated new node by artificial potential field method to perform local path planning, which specifically includes the following steps: S1: Calculate the attraction of the target point to the end effector of the industrial robot. Through the attraction, the new node will move toward the target point faster, and the path tends to the target, avoiding the problem of excessive randomness that may occur in the original RRT. The calculation formula of the attraction of the target point is as follows: F att =-K att (XX goal ) Where X represents the current position of the end effector of the industrial robot; X goal Indicates the target point position; K att represents the gravitational gain coefficient, XX goal Represents the Euclidean distance between the current industrial robot end effector and the center point of the obstacle; S2: Calculate the repulsive force of obstacles on the end effector of the industrial robot. By applying repulsive force to each obstacle, the new node will automatically avoid the obstacle, making the path smoother and avoiding collision with the obstacle. The calculation formula for the repulsive force of the obstacle is as follows: Where X represents the current position of the end effector of the industrial robot; X obs Indicates the position of the obstacle center; K rcp Represents the repulsion gain coefficient; P represents the distance of the obstacle influence range; XX obs Represents the Euclidean distance between the current industrial robot end effector and the center point of the obstacle; S3: Calculate the resultant force and adjust the new node position. Combine the attraction and repulsion forces to perform local path planning, and adjust the new node to move toward the target point while avoiding collision with obstacles. The calculation formulas for the resultant force and the new node position are as follows: F total =F att +F rcp In the formula, X new1 Indicates the original node position; Step indicates the iteration step; ||F total || represents the modulus of the resultant force.

[0021] Preferably, after the path is generated in step nine, a greedy strategy optimization is performed on it, and redundant nodes on the path are deleted to obtain an optimized path. The specific implementation method is as follows: Starting from the starting point of the path, check whether the path between each node and the subsequent node collides. If the path between two nodes does not collide, all intermediate nodes between the two nodes can be removed until a path collision occurs; if a collision occurs, start again from the previous node of the current node and continue to check the paths between the subsequent nodes until there are no redundant nodes in all paths, and the optimized path is obtained; through the above operations, all redundant nodes on the path can be removed, making the path more concise and effective.

[0022] Preferably, the step 10 of using a cubic B-spline curve to smooth the optimization path specifically includes the following contents: The nodes on the path after removing redundant points are used as the control points of the basis function of the B-spline curve, and the curve equation of the cubic B-spline can be obtained as follows: In the formula, c i F is the path node after removing redundant points. i,k (t) represents the basis function of cubic B-spline.

[0023] Beneficial effects of the present invention: The RRT-TAGB algorithm proposed in the present invention can quickly find obstacle avoidance paths in complex environments, greatly improves the guidance of random tree growth, greatly reduces redundant nodes on the path, shortens the search time, and makes the final path smoother. The overall obstacle avoidance path quality is higher, ensuring the adaptability of industrial robots to grasp and avoid obstacles in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0025] Figure 2 Schematic diagram of the principle of local path planning using the artificial potential field method in an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of the principle of removing redundant nodes using a greedy strategy in an embodiment of the present invention;

[0027] Figure 4 It is an obstacle map of the working space grasped by the end effector of the industrial robot in an embodiment of the present invention;

[0028] Figure 5 is a path diagram obtained by a traditional RRT algorithm in an embodiment of the present invention;

[0029] Figure 6 This is a path diagram obtained by the RRT-TAGB algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] like Figure 1 The figure shows a schematic diagram of the overall process of the present invention, which specifically includes ten steps: step 1, using a depth camera to obtain the grasping operation workspace of the industrial robot; step 2, simplifying the obstacles in the workspace into spherical obstacles and rectangular obstacles, and establishing an obstacle map; step 3, initializing variables, determining the starting point, target point and search step; step 4, according to the obstacle information of the workspace, determining the adaptive target bias probability value, and performing adaptive target bias sampling; step 5, iterating the random tree, selecting a closest node from the tree, and then generating a new node by sampling; step 6, local path planning, adjusting the position of the initially generated new node by the artificial potential field method, and performing local path planning; step 7, collision detection, checking whether there is an obstacle between the new node and the nearest node on the tree. If there is no collision, add the new node to the tree; if a collision occurs, return to step five and resample; step eight, path update, add the new node without collision to the tree, if the distance between the new node and the end point is smaller than the threshold, it is considered that the path has been found and the loop is exited; step nine, remove redundant nodes, generate a path, optimize it with a greedy algorithm, delete the redundant nodes on the path, and obtain the optimized path; step ten, use a cubic B-spline curve to smooth the optimized path to obtain the final industrial robot's grasping and obstacle avoidance path.

[0032] like Figure 2 FIG. 1 is a schematic diagram showing the principle of local path planning using the artificial potential field method in an embodiment of the present invention; FIG. Figure 2 It can be seen that the attraction of the target point enables the node to move towards the target, and the repulsion of the obstacle enables the node to avoid obstacles, which makes the planned path smoother and safer, avoiding the common broken line and obstacle collision problems in the traditional RRT algorithm; Figure 2 "X new ” is the position of the new node after the local optimization path.

[0033] like Figure 3 A schematic diagram of the principle of removing redundant nodes using a greedy strategy in an embodiment of the present invention; Figure 3 In the figure, the dotted line represents the original path and the solid line represents the pruned path. The path is optimized by gradually checking whether the middle point can be removed for every three consecutive points. During each check, if there is no obstacle blocking the connection of the straight line segment, the redundant points in the middle can be deleted to reduce the complexity of the path. The path finally returned is the optimized path, which contains the least points, and the straight line segment between every two points does not collide with obstacles. After removing all redundant nodes on the path through the greedy strategy, the path is more concise and effective.

[0034] In order to fully understand the technical solution of the present invention, the superiority of the RRT-TAGB algorithm is further demonstrated through simulation experiments. The following are the relevant parameter settings of the simulation experiments: The configuration environment of this simulation experiment uses a 64-bit Windows 10 operating system, 8GB of running memory, a graphics card driver of NVIDIA GeForce GTX1050Ti, and an experimental platform of Matlab2018b. The gripping working environment space of the industrial robot end is set to 100cm×100cm× 100cm, the starting point coordinates are (0, 0, 0), the target point coordinates are (100, 100, 100), and other relevant experimental parameters are shown in Table 1 below: Table 1 Related simulation experiment parameters like Figure 4 The figure shows an obstacle map of the grasping operation space of the industrial robot end effector established in an embodiment of the present invention; the black sphere and the green rectangular area represent simplified obstacles, i.e., inaccessible areas, and the white area represents an open space, i.e., a passable area, and the related spatial coordinate information is shown in Table 4; Figure 4 The obstacle map shown represents a simplified result of the real environment of the industrial robot grasping work space acquired using a depth camera. Table 2 Obstacle related parameter information

[0035] like Figure 5 The figure shows a path diagram obtained by a traditional RRT algorithm in an embodiment of the present invention; Figure 5 In the figure, the red curve is the final obstacle avoidance path, and the blue curve represents the search process. As can be seen from the figure, the traditional RRT algorithm generates more redundant nodes due to the randomness of tree growth, and the planned path has many turning points and is not smooth enough.

[0036] like Figure 6 Shown is a path diagram obtained by the RRT-TAGB algorithm in an embodiment of the present invention. Figure 6 (a) and (b) are the path comparison diagrams before and after smoothing. It can be seen from the figure that the path obtained by the RRT-TAGB algorithm has fewer redundant nodes. After smoothing, the path smoothness is significantly improved.

[0037] In order to avoid the contingency of the above results, 20 repeated experiments were conducted on the RRT and RRT-TAGB algorithms in an obstacle environment, and the average value was taken as the experimental evaluation index. The results are shown in Table 3 below: Table 3. Simulation results of obstacle avoidance path planning It can be seen from the table that compared with the traditional RRT algorithm, the average planning time of the RRT-TAGB algorithm is reduced by 72.2%, the average path length is reduced by 21.5%, the average number of path nodes is reduced by 88.2%, and the average number of iterations is reduced by 77.9%.

[0038] The description of the above embodiments is to facilitate the understanding of the present invention by those skilled in the art. Those skilled in the art to which the present invention belongs can still modify or replace the specific real-time mode of the present invention with equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. The present invention discloses an industrial robot grasping obstacle avoidance path planning method based on RRT-TAGB algorithm. It is characterized in that The specific steps include: Step 1: Use a depth camera to obtain the grasping workspace of the industrial robot; Step 2: Simplify the obstacles in the workspace into spherical obstacles and rectangular obstacles, and establish an obstacle map; Step 3: Initialize variables, determine the starting point, target point and search step size; Step 4: Determine the adaptive target bias probability value according to the obstacle information of the workspace, and perform adaptive target bias sampling; Step 5: Iterate the random tree, select a closest node from the tree, and then generate a new node by sampling; Step 6: Local path planning: adjust the position of the newly generated nodes through the artificial potential field method. Perform local path planning; Step 7: Collision detection: check if there is an obstacle between the new node and the nearest node in the tree. If there is no collision, add the new node to the tree; If a collision occurs, return to step 5 and resample; Step 8: Update the path. Add new nodes without collision to the tree. If the distance between the new node and the end point is smaller than the threshold, it is considered that the path has been found and the loop is exited. Step 9: Remove redundant nodes. After the path is generated, perform greedy strategy optimization on it, delete the redundant nodes on the path, and obtain the optimized path; Step 10: Use the cubic B-spline curve to smooth the optimized path to obtain the final grasping and obstacle avoidance path of the industrial robot.

2. The method for grasping and avoiding obstacles of an industrial robot based on the RRT-TAGB algorithm according to claim 1 is characterized in that: In step 4, the adaptive target bias probability value is determined by the following method: Where λ represents the obstacle volume correction coefficient, which is used to improve the error between the obstacle equivalent model volume and the actual volume, and its value is 0.8~1.2; V c Represents the volume of the rectangular obstacle; V q Represents the volume of the spherical obstacle; V k Represents the volume of the working space at the end of an industrial robot.

3. The method for grasping and avoiding obstacles of an industrial robot based on the RRT-TAGB algorithm according to claim 1 is characterized in that: In step 4, adaptive target bias sampling is performed, and the specific implementation process is as follows: According to the adaptive target bias probability P target (value range is 0~1), each time a node is expanded, a new probability value P is uniformly and randomly generated between 0 and 1; when P>P target When , the sampling point is still randomly selected in the workspace; otherwise, the target point X is directly selected. goal as the sampling point.

4. The method for grasping and avoiding obstacles of an industrial robot based on the RRT-TAGB algorithm according to claim 1 is characterized in that In step 6, the newly generated nodes are initially adjusted in position by artificial potential field method to perform local path planning. The specific implementation process is as follows: S1: Calculate the attraction of the target point to the end effector of the industrial robot. Through the attraction, the node will move toward the target point faster, and the path tends to the target, avoiding the problem of excessive randomness that may occur in the original RRT; S2: Calculate the repulsive force of obstacles on the end effector of the industrial robot. By applying repulsive force to each obstacle, the new node will automatically avoid the obstacle, making the path smoother and avoiding collision with obstacles. S3: Calculate the resultant force and adjust the position of the new node. Combine the attractive force and the repulsive force to perform local path planning. Adjust the new node to move toward the target point while avoiding collision with obstacles.

5. The method for grasping and avoiding obstacles of an industrial robot based on the RRT-TAGB algorithm according to claim 1 is characterized in that The redundant nodes are removed in step nine, specifically including the following steps: Starting from the starting point of the path, check whether the path between each node and the subsequent node collides in turn; if the path between two nodes does not collide, all intermediate nodes between the two nodes can be removed until a path collision is encountered; if a collision occurs, start again from the previous node of the current node, and continue to check the paths between the subsequent nodes until there are no redundant nodes in all paths; the path finally returned is the optimized path, which contains the least nodes, and there is no collision between any two nodes and obstacles.

6. According to the method for grasping and avoiding obstacles of an industrial robot based on the RRT-TAGB algorithm of claim 1, in step 10, a cubic B-spline curve is used to smooth the path, and the specific implementation process is as follows: Using the nodes on the path after removing redundant points as the control points of the basis function of the B-spline curve, the curve equation of the cubic B-spline can be obtained as follows: In the formula, c i F is the path node after removing redundant points. i,k (t) represents the basis function of cubic B-spline.

Images