A robot flexible welding path planning method for non-standard workpieces

Abstract

The present invention discloses a robot flexible welding path planning method for non-standard workpieces. The present invention comprises the following steps: first, after acquiring and processing the surface point cloud information of the welding workpiece, a three-dimensional model of the welding workpiece is obtained; then, a feature recognition algorithm is used to complete the weld extraction; then, the weld posture and the welding gun posture are determined; a meta-heuristic algorithm is applied to determine the welding sequence of the weld; finally, an efficient path planning algorithm is applied to quickly find a collision-free path to complete the welding path planning; and the planned welding path parameters are used to guide the welding robot to perform welding. The present invention is aimed at multi-specification small-batch workpieces, combined with data processing methods such as point cloud three-dimensional reconstruction and weld feature extraction, and an efficient path planning method is applied to quickly find a collision-free path for the welding robot, which can effectively improve the path quality and compress the planning time, while ensuring the continuity of the welding operation and the consistency of the welding quality.

Description

Technical Field

[0001] The present invention belongs to a robot flexible welding path planning method in the technical field of robot flexible welding, and particularly relates to a robot flexible welding path planning method for non-standard (i.e., multi-specification and small-batch) workpieces. Background Technique

[0002] With the in-depth development of industrial automation, industrial robots have been widely used in standardized and streamlined operations in industries such as the automotive industry, electronic assembly, metallurgy and chemical industry, and food processing, mainly for operations such as grinding, painting, loading and unloading, and welding. At present, the welding tasks of large quantities of standardized products have been automated through manual teaching programming. However, in fields such as shipbuilding, offshore engineering, and steel structures, the production has a high degree of diversity and non-standardization, and it is difficult to compile a unified welding program to meet the flexible welding requirements of multi-specification and small-batch workpieces. Traditional manual teaching methods have many drawbacks, such as a long design cycle, low production efficiency, and high labor costs, etc. Therefore, for such personalized, multi-specification, and small-batch workpieces, it is necessary to introduce robot flexible welding technology and develop a more automated and intelligent robot welding method.

[0003] Robot flexible welding is an advanced welding technology that uses a multi-axis industrial robot to operate, and can flexibly configure corresponding data processing algorithms according to different welding processes and requirements, extract processing information and control the robot to complete a series of welding tasks. This technology can significantly improve the automation level and efficiency of welding operations, and realize the automatic welding of high-efficiency, high-quality, high-precision, and flexible multi-specification and small-batch workpieces. The related research on robot flexible welding technology mainly focuses on multiple aspects such as weld start position guidance, weld extraction, weld tracking, welding pool monitoring, and welding path planning. Among them, the welding path planning technology is particularly crucial, requiring the robot to plan a collision-free and optimized torch movement path according to the recognized workpiece information and weld information. The existing flexible welding path planning methods mainly include:

[0004] 1) Path planning based on specific parameters. This method uses a predefined series of parameters, such as torch speed, torch angle, and torch-to-workpiece distance, etc., and generates a path suitable for a specific welding task by adjusting these parameters. This method lacks flexibility. Especially when dealing with complex or non-standard workpieces, it is difficult to adjust the parameters and it is difficult to meet the welding requirements.

[0005] 2) Path planning based on sensor feedback. When forming the welding path, this method relies on the real-time feedback of the sensor for dynamic optimization and adjustment. When dealing with complex or non-standard workpieces, this method can significantly improve the adaptability of the welding process and the welding accuracy. However, this method highly depends on the sensor accuracy and response speed, has high requirements for the welding environment, and when facing frequently changing workpiece types and welding requirements, it is necessary to continuously adjust and calibrate the sensor system, resulting in high costs.

[0006] 3) Path planning based on rasterized maps. This method first divides the welding area into grids to establish a rasterized map, and then uses graph search algorithms such as the D*Lit algorithm, Theta* algorithm, Dijkstra algorithm, and A* algorithm on this map to find the solution path. The accuracy of the path planning of this method is limited by the resolution of the grid. Therefore, it takes a high time cost and computational cost to establish the rasterized map, which is not conducive to application in multi-specification and small-batch workpieces with rapidly changing production requirements. Summary of the Invention

[0007] In the process of achieving welding automation for multi-specification and small-batch workpieces, in order to solve the defects existing in the background technology, the present invention provides a robot flexible welding path planning method for multi-specification and small-batch workpieces. The present invention can effectively reduce the path cost of the planning, plan a simple and efficient optimized path for the robot, and has the characteristics of intelligence, teaching-free, high efficiency, and high flexibility.

[0008] The technical solution of the present invention is as follows:

[0009] Step 1: After acquiring and processing the surface point cloud information of the welding workpiece, obtain the three-dimensional model of the welding workpiece;

[0010] Step 2: Based on the dimensional information of the welding workpiece, extract the weld seams of the three-dimensional model of the workpiece to obtain the weld seams of the welding workpiece;

[0011] Step 3: Based on the three-dimensional model and weld seams of the welding workpiece, determine the weld seam posture and the welding torch posture;

[0012] Step 4: According to the weld seams of the welding workpiece, determine the welding sequence of the weld seams;

[0013] Step 5: Use the method of modeling with oriented bounding boxes to determine the obstacle objects, and then establish a welding environment model;

[0014] Step 6: Based on the weld seam posture, welding torch posture, and welding environment model, combined with the welding sequence of the weld seams, apply an efficient path planning algorithm to quickly find a collision-free path and obtain a welding path;

[0015] Step 7: Send the welding path parameters to the robot control cabinet to guide the welding robot to weld the welding workpiece.

[0016] In step 4, according to the weld seam of the welded workpiece, a meta-heuristic algorithm is used to determine the welding sequence of the weld seam.

[0017] In step 6, find the projection planes corresponding to the start and end points of the weld seam of the workpiece in the bounding box, determine the projection points corresponding to the start and end points of the weld seam in the projection plane, and use a simple path planning algorithm to calculate the planned path between the start state point of the robot and the projection point corresponding to the start end point of the weld seam, and the planned path between the projection point corresponding to the end point of the weld seam and the stop state point of the robot, and use an advanced path planning algorithm to calculate the planned path between the projection point corresponding to the start end point of the weld seam and the start end point of the weld seam, the planned path between the end point of the weld seam and the projection point corresponding to the end point of the weld seam, and the planned path between two adjacent separated weld seams, so as to obtain the welding planned path.

[0018] For the planned path between the start state point of the robot and the projection point corresponding to the start end point of the weld seam, the specific calculation process is as follows:

[0019] First, use the fifth-order polynomial interpolation method to generate the path between the start state point of the robot and the projection point corresponding to the start end point of the weld seam, then discretize the path and use the separating axis theorem to perform collision detection between the discretized path and the oriented bounding box and locally adjust the path to obtain the planned path between the start state point of the robot and the projection point corresponding to the start end point of the weld seam.

[0020] The specific steps for calculating the planned path using the advanced path planning algorithm are as follows:

[0021] S1: Establish a robot path graph A and a robot path graph B, with the starting state x of the robotic arm start as the initial node of the robot path graph A, and the target state x of the robotic arm end as the initial node of the robot path graph B; among them, the starting state x of the robotic arm start is the projection point corresponding to the start end point of the weld seam, the end point of the weld seam or the starting point of two adjacent separated weld seams; the target state x of the robotic arm end is the start end point of the weld seam, the projection point corresponding to the end point of the weld seam or the end point of two adjacent separated weld seams;

[0022] S2: Based on the weld seam pose, the welding torch pose and the welding environment model, alternately expand and optimize the robot path graph A and the robot path graph B until the two graphs coincide at a node, and backtrack from this node to the starting state x of the robotic arm start and the target state x of the robotic arm end to obtain the initial planned path;

[0023] S3: Perform dynamic optimization based on the initial planned path to obtain the optimal planned path.

[0024] Specifically, S2 is as follows:

[0025] S2.1: Take the joint space of the robotic arm as the initial sampling space, and randomly generate a sampling node x in the current sampling space new , introduce the target bias probability P, and with probability P, the sampling node x new is a random node in the robot path graph B;

[0026] S2.2: Traverse the distances between all nodes in the robot path graph A and the current sampling node x new . When the distance between the closest node to the current sampling node x new and x new ,x closestinA ) is less than or equal to the node limit distance S limit , then enter S2.3; otherwise, move the closest node x new to the current sampling node x closestinA by the node limit distance S limit to a new node, update this new node as the current sampling node x new , and then enter S2.3;

[0027] S2.3: Perform collision detection on the current sampling node x new using the three-dimensional model of the welded workpiece. If the collision detection is passed, include the current sampling node x new into the robot path graph A; otherwise, execute S2.1 - S2.2 again until the current sampling node x new passes the collision detection and is placed in the robot path graph A;

[0028] S2.4: In the robot path graph A, perform partial graph reorganization on the current sampling node x new according to the path cost to obtain the node order of the robot path graph A;

[0029] S2.5: Traverse the distances between all nodes in the robot path graph B and the current sampling node x new . When the distance between the closest node to the current sampling node x new and x new ,x closestinB ) is less than or equal to the node limit distance S limit , then enter S2.7; otherwise, move the closest node x new to the current sampling node x closestinB by the node limit distance S limitTo a new node, update the new node as the current sampled node x new , and then enter S2.6;

[0030] S2.6: Perform collision detection on the current sampled node x using the three-dimensional model of the welded workpiece new . If the collision detection is passed, include the current sampled node x new in the robot path map B. Otherwise, regenerate the sampled node x new . The sampled node x new has a probability of P to be a random node in the robot path map A until the current sampled node x new and the distance dis(x new ,x closestinB ) between the nearest node in the robot path map B is less than or equal to the node limit distance S limit and passes the collision detection, then put it into the robot path map B. Then, in the robot path map B, perform partial graph reorganization on the current sampled node x new according to the path cost to obtain the node order of the robot path map B;

[0031] S2.7: Repeat S2.1 to S2.6 to obtain the final node orders in the robot path maps A and B. Then, start backtracking from the current sampled node x new to obtain the initial planned path.

[0032] The partial graph reorganization of the current sampled node x according to the path cost is specifically as follows: new In the robot path map A or the robot path map B, for each node x

[0033] in the neighborhood N of the current sampled node x new , calculate the new path cost of connecting to the neighborhood node x neighbor through the current sampled node x new . If it is less than the current path cost of the neighborhood node x neighbor , then replace the current sampled node x neighbor with the superior node of the neighborhood node x bew . Then calculate the new path cost of connecting to the current sampled node x neighbor through the neighborhood node x neighbor . If it is less than the current path cost of the current sampled node x new , then replace the neighborhood node x new with the superior node of the current sampled node x neighbor . new The target bias probability P is dynamically adjusted according to the heuristic criterion. The formula is:

[0034] ​

[0035]

[0036] Among them, P(i) represents the target bias probability at the current iteration number i, and i mid represents the expected midpoint iteration number, and α represents a positive scaling factor.

[0037] The formula for the path cost is as follows:

[0038] C = ω 1 C length + ω 2 C collision + ω 3 C joint_limits + ω 4 C stability + ω 5 C smoothness

[0039] Among them, C represents the path cost, C length represents the path length, C collision represents the collision risk, C joint_limits represents the robot joint limit, C stability represents the robot pose stability, C smoothnes represents the motion smoothness; ω 1 ~ω 5 represent the first to fifth weights.

[0040] Specifically, S3 is as follows:

[0041] S3.1: Based on the initial planned path, construct a measurement matrix M. After performing singular value decomposition and calculation on the measurement matrix M, obtain the hyperellipsoid rotation matrix C. The formula is as follows:

[0042]

[0043]

[0044] U∑V T ≡ M

[0045] C = Udiag{1,...,det(U)det(V)}V T

[0046] Among them, a 1 represents the direction from the starting point x start to the ending point x end of the initial planned path. U, ∑, and V T are the left singular vector matrix, diagonal matrix, and right singular vector transpose matrix respectively, and |||| 2 represents the Euclidean norm; I 1is a 1*n dimensional identity matrix, T represents transpose, det() represents the determinant of a matrix; diag{} represents a diagonal matrix, and the elements inside {} are the elements on the diagonal of the matrix;

[0047] S3.2: According to the length of the current planned path, the starting point x start and the end point x end establish a super-ellipsoid, and the formulas for the relevant geometric parameters are as follows:

[0048]

[0049]

[0050]

[0051] Among them, a, b, and c respectively represent the semi-major axis, semi-minor axis, and focal length of the super-ellipsoid. c best represents the length of the current planned path, and c min represents the distance between the starting point x start of the current planned path and the end point x end ;

[0052] S3.3: Construct the shape matrix L of the super-ellipsoid, and the formula is as follows:

[0053]

[0054] S3.4: Randomly sample a node x ball inside the unit hypersphere. After scaling, rotating, and translating the node x ball by the shape matrix L and the super-ellipsoid rotation matrix C, obtain the sampling point x' new inside the super-ellipsoid;

[0055] S3.5: Traverse the distances between all nodes in the robot path map A and the robot path map B and the current sampling point x' new . When the distance dis(x' new , x closet ) corresponding to the node x new that is closest to the current sampling point x' closest is less than or equal to the node limit distance S limit , then enter S3.6; otherwise, move the node x new that is closest to the current sampling point x' closet along the third sampling direction by the node limit distance S limit to a new node, update this new node as the sampling point x' new , and then enter S3.6;

[0056] S3.6: Use the three-dimensional model of the welded workpiece for the current sampling point x'new Perform collision detection. If the collision detection is passed, the current sampling point x' new is incorporated into the robot path map A or the robot path B. Otherwise, S3.4 - S3.5 are executed again until the current sampling point x' new passes the collision detection and is placed in the robot path map A or the robot path B;

[0057] S3.7: Dynamically adjust the size of the neighborhood N based on the cost function. The formula is as follows:

[0058]

[0059] where N size represents the adjusted neighborhood size, N D represents the original neighborhood size, L current represents the current optimal path length, L baseline represents the reference path length, and β represents an adjustable dynamic parameter for controlling the rate;

[0060] S3.8: Based on the current neighborhood N size, perform partial graph reorganization on the current sampling node x' new to obtain a new node order in the robot path map A or the robot path map B;

[0061] S3.9: Repeat S3.2 - S3.8 to obtain the optimal planned path.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. The present invention can flexibly configure corresponding data processing algorithms according to different welding processes and requirements, extract processing information, and control the robot to complete a series of welding tasks. It is universal for workpieces with multiple specifications and small batches, and solves the problems of insufficient flexibility and low production efficiency in the prior art when dealing with diversified and non-standard products.

[0064] 2. The present invention proposes an efficient path planning method, including a simple path planning algorithm and an advanced path planning algorithm. Combining the two algorithms to quickly find a collision-free path for the welding robot can effectively improve the path quality and compress the planning time, while ensuring the continuity of the welding operation and the consistency of the welding quality.

[0065] 3. The present invention proposes a robot flexible welding method for workpieces with multiple specifications and small batches. Before applying the efficient path planning method proposed by the invention, advanced automation technologies and data processing methods such as point cloud three-dimensional reconstruction and weld seam feature extraction are effectively integrated, greatly improving the automation and integration degree of the welding process, and promoting the development of intelligent manufacturing. Description of the Drawings

[0066] Figure 1 This is the overall flowchart of the robot flexible welding path planning method provided by the embodiments of the present invention.

[0067] Figure 2 It is an automatic welding system for multi-specification and small-batch workpieces.

[0068] Figure 3 This is the flowchart for obtaining the initial path of the alternating expansion and optimization path diagram in the embodiments of the present invention.

[0069] Figure 4 This is the flowchart for dynamically optimizing the initial path to obtain the optimal path in the embodiments of the present invention.

[0070] In the figure: 1. Host computer; 2. Welding power source; 3. Robot control cabinet; 4. Welding robot; 5. Turntable; 6. Welding workpiece; 7. Scanning frame; 8. Three-dimensional structured light camera. Specific implementation manners

[0071] The following further describes the present invention in detail with reference to the drawings and embodiments. It should be noted that the following embodiments are intended to facilitate the understanding of the present invention and do not limit it in any way.

[0072] As shown in the Figure 2 accompanying drawings, an automatic welding system for multi-specification and small-batch workpieces includes: a host computer 1, a welding power source 2, a robot control cabinet 3, a welding robot 4, a turntable 5, a welding workpiece 6, a scanning frame 7, and a three-dimensional structured light camera 8.

[0073] The host computer 1 is used to receive the workpiece surface point cloud information captured by the three-dimensional structured light camera 8, perform point cloud stitching and workpiece surface reconstruction, complete weld seam extraction and welding path planning for the welding robot 4. The welding power source 2 provides power for the three-dimensional structured light camera 8 on the robot control cabinet 3, the welding robot 4, the turntable 5, and the scanning frame 7. The robot control cabinet 3 receives the welding path, trajectory parameters, and welding process parameters sent by the host computer 1, and controls the welding process of the welding robot 4 and the mechanism movement of the turntable 5. The welding workpiece 6 is placed on the turntable 5, and the height and angle of the camera can be adjusted by adjusting the position of the hinge.

[0074] The present invention proposes an efficient path planning algorithm, enabling the robot to quickly find the optimal collision-free path;

[0075] As Figure 1 shown, a robot flexible welding path planning method for multi-specification and small-batch workpieces includes the following steps:

[0076] Step 1: After obtaining and processing the point cloud information of the surface of the welding workpiece, specifically, the point cloud is stitched and the three-dimensional reconstruction of the workpiece surface is performed to obtain the three-dimensional model of the welding workpiece. In this embodiment, the turntable 5 cooperates with the three-dimensional structured light camera 8 to capture the surface point cloud information of the welding workpiece 6, and transmits the information to the host computer 1. The host computer 1 receives the surface point cloud information of the workpiece captured by the three-dimensional structured light camera 8, and then performs point cloud stitching and three-dimensional reconstruction of the workpiece surface to obtain the three-dimensional model of the welding workpiece.

[0077] Step 2: Based on the dimensional information of the welding workpiece, the welds of the three-dimensional model of the workpiece are extracted by using a feature recognition algorithm to obtain the welds of the welding workpiece.

[0078] Step 3: Based on the three-dimensional model and welds of the welding workpiece, determine the weld pose and the welding torch pose. The weld pose is the spatial orientation of the weld pose, and the welding torch pose is the spatial orientation of the welding torch pose.

[0079] Step 4: According to the welds of the welding workpiece, apply a meta-heuristic algorithm to determine the welding sequence of the welds.

[0080] Step 5: Use the method of modeling with an oriented bounding box to determine the obstacle objects, and then establish a welding environment model.

[0081] Step 6: Based on the weld pose, the welding torch pose, and the welding environment model (such as the three-dimensional point cloud coordinate information of the obstacles), combined with the welding sequence of the welds, apply an efficient path planning algorithm to quickly find a collision-free path to obtain the welding path.

[0082] In Step 6, find the projection planes corresponding to the start and end points of the welds of the workpiece in the bounding box, and determine the projection points corresponding to the start and end points of the welds in the projection planes. Specifically: for each surface of the oriented bounding box except the bottom surface, calculate the normal vector of each surface by using the cross product of two vectors, determine the plane equation of each surface in combination with the center point and half length of the bounding box, calculate the distances from the start and end points of the welds to each surface according to the point-to-plane distance formula, find the surface with the shortest distance respectively, move the start and end points of the welds along the normal vector of the surface to obtain their projection points on the surface, and use the projection points as the intermediate points of the paths between the start and end points of the welds and the start or stop motion states of the robot. Use a simple path planning algorithm to calculate the planned paths between the start state point of the robot and the projection point corresponding to the start end point of the weld, and between the projection point corresponding to the end point of the weld and the stop state point of the robot, and use an advanced path planning algorithm to calculate the planned paths between the projection point corresponding to the start end point of the weld and the start end point of the weld, between the end point of the weld and the projection point corresponding to the end point of the weld, and between adjacent separated welds, so as to obtain the welding planned path. The start and end points of the welds are specifically the start point of the first weld and the end point of the last weld.

[0083] For the planned path between the starting state point of the robot and the projection point corresponding to the starting endpoint of the weld, the specific calculation process is as follows:

[0084] First, use the fifth-degree polynomial interpolation method to generate a smooth path between the starting state point of the robot and the projection point corresponding to the starting endpoint of the weld. Then, after discretizing this path, use the separating axis theorem to perform collision detection between the discretized path and the oriented bounding box and locally adjust the path to obtain the planned path between the starting state point of the robot and the projection point corresponding to the starting endpoint of the weld. The calculation process for the planned path between the projection point corresponding to the ending endpoint of the weld and the stopping state point of the robot is the same.

[0085] The specific steps for calculating the planned path using an advanced path planning algorithm are as follows:

[0086] S1: Establish the robot path graph A and the robot path graph B. Take the starting state x of the robotic arm start as the initial node of the robot path graph A, and take the target state x of the robotic arm end as the initial node of the robot path graph B. Determine the number of iterations and the limit step size S limit ;

[0087] In S1, the starting state x of the robotic arm start is the projection point corresponding to the starting endpoint of the weld, the ending endpoint of the weld, or the starting point of two adjacent separated welds; the target state x of the robotic arm end is the starting endpoint of the weld, the projection point corresponding to the ending endpoint of the weld, or the ending point of two adjacent separated welds.

[0088] S2: Based on the weld pose, the torch pose, and the welding environment model, alternately expand and optimize the robot path graph A and the robot path graph B until the two graphs coincide at a node. After backtracking from this node to the starting state x of the robotic arm start and the target state x of the robotic arm end , obtain the initial planned path.

[0089] As Figure 3 shown, S2 is specifically as follows:

[0090] S2.1: Take the joint space of the robotic arm as the initial sampling space, randomly generate a sampling node x new in the current sampling space, introduce the target bias probability P. The sampling node x new has a probability of P to be a random node in the robot path graph B and a probability of 1 - P to be a random node in the joint space;

[0091] The target bias probability P is dynamically adjusted according to the heuristic criterion. The formula is:

[0092]

[0093] Where P(i) represents the target paranoia probability at the current iteration number i, i mid represents the expected number of midpoint iterations, and α represents a positive scaling factor.

[0094] S2.2: Traverse all nodes in the robot path graph A and the current sampling node x new The distance between the current sampling node x new The distance between the nearest nodes dis(x new ,x closestinA ) is less than or equal to the node limit distance S limit , then enter S2.3; otherwise, along the first sampling direction, it will be connected with the current sampling node x new Nearest node x closestinA Mobile node limit distance S limit Go to a new node and update the new node to the current sampling node x new , then enter S2.3;

[0095] The first sampling direction is a unit vector direction, the formula is:

[0096]

[0097] S2.3: Using the 3D model of the welding workpiece, that is, the workpiece point cloud data after 3D reconstruction, the current sampling node x new Perform collision detection. If the collision detection passes, the current sampling node x new Incorporate it into the robot path graph A, otherwise execute S2.1~S2.2 again until the current sampling node x new Through collision detection and placed in the robot path map A;

[0098] S2.4: In the robot path graph A, the current sampling node x is calculated according to the path cost new Reorganize the partial graph to obtain the node order of the robot path graph A;

[0099] S2.5: Traverse all nodes in the robot path graph B and the current sampling node x new The distance between the current sampling node x new The distance between the nearest nodes dis(x new ,x closestinB ) is less than or equal to the node limit distance S limit , that is, the current sampling node x in the robot path graph A new The distance to the nearest node in the robot path graph B is less than or equal to the node limit distance S limit, then enter S2.7; otherwise, move the node closest to the current sampling node x new along the second sampling direction by a restricted distance S closestinB to a new node, update the new node as the current sampling node x limit , and then enter S2.6; new

[0100] The second sampling direction is the direction of the unit vector , and the formula is:

[0101]

[0102]

[0103] S2.6: Use the three-dimensional model of the welded workpiece, that is, the point cloud data of the workpiece after three-dimensional reconstruction, to perform collision detection on the current sampling node x new through the PCL point cloud collision detection library. If the collision detection is passed, include the current sampling node x new in the robot path graph B; otherwise, regenerate the sampling node x new . The sampling node x new has a probability of P to be a random node in the robot path graph A and a probability of 1 - P to be a random node in the joint space until the distance dis(x new , x new closestinB ) between the current sampling node x and the closest node in the robot path graph B is less than or equal to the node restricted distance S limit and the collision detection is passed, then put it into the robot path graph B. Then, in the robot path graph B, perform partial graph reorganization on the current sampling node x new according to the path cost to obtain the node order of the robot path graph B;

[0104] Perform partial graph reorganization on the current sampling node x new according to the path cost, specifically:

[0104] In the robot path graph A or the robot path graph B, for each node x new in the neighborhood N of the current sampling node x neighbor , calculate the new path cost of connecting to the neighborhood node x new through the current sampling node x neighbor . If it is less than the current path cost of the neighborhood node x neighbor , then replace the current sampling node x new with the superior node of the neighborhood node x neighbor ; then calculate the new path cost of connecting to the current sampling node x neighbor through the neighborhood node x new . If it is less than the current path cost of the current sampling node x new , then replace the neighborhood node xneighbor is the upper-level node of the current sampling node x new ;

[0105] Among them, the neighborhood N is a hypersphere neighborhood, which contains all the nodes in the path graph whose distance from the current sampling node x new does not exceed the distance radius D. The formula is:

[0106] N(x new ) = {x in the joint space: dis(x, x new ) ≤ D}

[0107] The formula for the path cost is as follows:

[0108] C = ω 1 C length + ω 2 C collision + ω 3 C joint_limits + ω 4 C stability + ω 5 C smoothness

[0109] Among them, C represents the path cost, C len represents the path length, C collision represents the collision risk, C joint_limits represents the robot joint limit, C stability represents the robot pose stability, C smoothness represents the motion smoothness; ω 1 ~ω 5 represent the first to fifth weights.

[0110] S2.7: Repeat S2.1 to S2.6 to obtain the node order in the final robot path graph A and robot path graph B, and then perform backtracking starting from the current sampling node x new to obtain the initial planned path.

[0111] S3: Perform dynamic optimization based on the initial planned path to obtain the optimal planned path.

[0112] As Figure 4 shown, S3 is specifically:

[0113] S3.1: Construct a measurement matrix M based on the initial planned path. After performing singular value decomposition and calculation on the measurement matrix M, obtain the hyperellipsoid rotation matrix C. The formula is as follows:

[0114]

[0115]

[0116] U∑V T ≡M

[0117] C = U diag{1,..., det(U) det(V)} V T

[0118] where a 1 is a unit vector representing the direction from the starting point x start to the ending point x end of the initial planned path, and U, ∑, V T represent the hyperellipsoid transformation matrices, which are the left singular vector matrix, the diagonal matrix, and the transpose of the right singular vector matrix respectively, || || 2 denotes the Euclidean norm; I 1 is a 1 * n dimensional identity matrix, T represents transpose, det( ) represents the determinant of a matrix; diag{} represents a diagonal matrix, and the elements inside {} are the elements on the diagonal of the matrix;

[0119] S3.2: Establish a hyperellipsoid based on the length of the current planned path, the starting point x start and the ending point x end . The formulas for the relevant geometric parameters are as follows:

[0120]

[0121]

[0122]

[0123] where a, b, and c represent the semi - major axis, semi - minor axis, and focal length of the hyperellipsoid respectively, c best represents the length of the current best path, c min represents the distance between the starting point x start and the ending point x end of the current planned path;

[0124] S3.3: Construct the shape matrix L of the hyperellipsoid. The shape matrix L is a diagonal matrix, and the elements on the diagonal are defined based on c best and c min , representing the axis lengths of the hyperellipsoid in each dimension and determining the shape of the hyperellipsoid in the high - dimensional space. The formula is as follows:

[0125]

[0126] S3.4: Randomly sample a node x ball inside the unit hypersphere, and obtain the sampling point x' inside the hyperellipsoid after scaling, rotating, and translating the node x ball by the shape matrix L and the hyperellipsoid rotation matrix Cnew ;

[0127] Sampling point x' new The acquisition formula is:

[0128] x ball ~u(x ball )

[0129] x new = CLx ball + x centre

[0130] where u() represents a function for random sampling within the unit hypersphere, and x centre represents the center point of the super-ellipse, that is, the midpoint of the two foci.

[0131] S3.5: Traverse all the nodes in the robot path map A and the robot path map B and the current sampling point x' new The distance between them. When the node x new corresponding to the closest distance to the current sampling point x' closet The corresponding distance dis(x' new , x closest ) is less than or equal to the node limit distance S limit , then enter S3.6; otherwise, move the node x new corresponding to the closest distance to the current sampling point x' closet by the node limit distance S limit to a new node, update this new node as the sampling point x' new , and then enter S3.6;

[0132] The second sampling direction is the direction of the unit vector . The formula is:

[0133]

[0134] S3.6: Use the three-dimensional model of the welded workpiece, that is, the point cloud data of the workpiece after three-dimensional reconstruction, to perform collision detection on the current sampling point x' new . If the collision detection is passed, then include the current sampling point x' new into the robot path map A or the robot path B. Otherwise, execute S3.4 - S3.5 again until the current sampling point x' new passes the collision detection and is placed in the robot path map A or the robot path B;

[0135] S3.7: Dynamically adjust the size of the neighborhood N based on the cost function. The formula is as follows:

[0136]

[0137] Among them, N size represents the adjusted domain size, N d represents the original domain size, L current represents the current best path length, L baseline represents the reference path length, and β represents an adjustable dynamic parameter for controlling the rate;

[0138] S3.8: Based on the current neighborhood N size, perform partial graph recombination on the current sampled node x' according to the path cost new to obtain a new node order in the robot path graph A or the robot path graph B;

[0139] S3.9: Repeat S3.2 - S3.8 until the set number of iterations or the solution cost is reached to obtain the optimal planned path.

[0140] Step Seven: Send the welding path parameters to the robot control cabinet to guide the welding robot to weld the welded workpiece.

[0141] Finally, it should be noted that the above embodiments and descriptions are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced. Without departing from the spirit and scope of the disclosure of the technical solutions of the present invention, they should all be covered by the protection scope of the claims of the present invention.

Claims

1. A robot flexible welding path planning method for non-standard workpieces, characterized in that: The following steps are involved: Step 1: After acquiring and processing the point cloud information of the welding workpiece surface, a three-dimensional model of the welding workpiece is obtained; Step 2: Based on the size information of the welding workpiece, the weld seam of the welding workpiece is extracted from the three-dimensional model of the workpiece to obtain the weld seam of the welding workpiece; Step 3: Determine the weld posture and welding gun posture based on the 3D model of the welding workpiece and the weld; Step 4: Determine the welding sequence of the welds according to the welds of the welded workpiece; Step 5: Use directed bounding box method to determine obstacle objects and then establish welding environment model; Step 6: Based on the weld posture, welding gun posture and welding environment model, combined with the welding sequence of the weld, an efficient path planning algorithm is applied to quickly find a collision-free path and obtain the welding path; In the step six, the projection surface corresponding to the start and end points of the weld of the workpiece is found in the bounding box, the projection point corresponding to the start and end points of the weld is determined in the projection surface, and the planning path between the robot start state point and the projection point corresponding to the weld start end point and the planning path between the projection point corresponding to the weld end end point and the robot stop state point are calculated by using a simple path planning algorithm, and the planning path between the projection point corresponding to the weld start end point and the weld start end point, the planning path between the weld end end point and the projection point corresponding to the weld end end point, and the planning path between two adjacent separated welds are calculated by using an advanced path planning algorithm, so as to obtain a welding planning path; The specific steps of calculating the planned path using the advanced path planning algorithm are as follows: S1: Create robot path graph A and robot path graph B, starting with the robot arm in state x start As the initial node of the robot path graph A, the robot arm target state x end As the initial node of the robot path graph B; where the robot arm starts in state x start is the projection point corresponding to the starting end point of the weld, the ending end point of the weld, or the starting point of two adjacent separated welds; the target state of the robot arm x end It is the projection point corresponding to the starting end point of the weld, the ending end point of the weld, or the end point of two adjacent separated welds; S2: Based on the weld posture, welding gun posture and welding environment model, alternately expand and optimize the robot path graph A and the robot path graph B until the two graphs overlap at a node, and then trace back from the node to the robot arm's starting state x start and the robot target state x end After that, the initial planning path is obtained; S3: Perform dynamic optimization based on the initial planning path to obtain the optimal planning path; The S2 includes: S2.1: Take the joint space of the robot as the initial sampling space and randomly generate a sampling node x in the current sampling space new , introduce the target paranoia probability P, sampling node x new There is a probability of P being a random node in the robot path graph B; The target paranoia probability P is dynamically adjusted according to the heuristic criterion, and the formula is: Among them, P(i) represents the target paranoia probability at the current iteration number i, i mid represents the expected number of midpoint iterations, and α represents a positive scaling factor; Step 7: Send the welding path parameters to the robot control cabinet to guide the welding robot to weld the workpiece.

2. A robot flexible welding path planning method for non-standard workpieces according to claim 1, characterized in that: In the step 4, a meta-heuristic algorithm is used to determine the welding sequence of the welds according to the welds of the welded workpiece.

3. The method for robot flexible welding path planning for non-standard workpieces according to claim 1, characterized in that: For the planned path between the robot's starting state point and the projection point corresponding to the weld start endpoint, the calculation process is specifically as follows: Firstly, the quintic polynomial interpolation method is used to generate the path between the starting state point of the robot and the projection point corresponding to the starting endpoint of the weld. Then, the path is discretized and the separating axis theorem is used to perform collision detection between the discretized path and the directed bounding box and locally adjust the path to obtain the planned path between the starting state point of the robot and the projection point corresponding to the starting endpoint of the weld.

4. The method for robot flexible welding path planning for non-standard workpieces according to claim 1, characterized in that: The S2 further includes: S2.2: Traverse all nodes in the robot path graph A and the current sampling node x new The distance between the current sampling node x new The distance between the closest nodes dis(x new ,x closestinA ) is less than or equal to the node limit distance S limit , then enter S2.3; otherwise, along the first sampling direction, it will be connected to the current sampling node x new The nearest node x closestinA Mobile node limit distance S limit Go to a new node and update the new node to the current sampling node x new , then enter S2.3; S2.3: Use the three-dimensional model of the welding workpiece to sample the current node x new Perform collision detection. If the collision detection passes, the current sampling node x new into the robot path graph A, otherwise execute S2.1 to S2.2 again until the current sampling node x new Through collision detection and put into the robot path map A; S2.4: In the robot path graph A, the current sampling node x is calculated according to the path cost. new Perform partial graph reorganization to obtain the node order of the robot path graph A; S2.5: Traverse all nodes in the robot path graph B and the current sampling node x new The distance between the current sampling node x new The distance between the closest nodes dis(x new ,x closestinB ) is less than or equal to the node limit distance S limit , then enter S2.7; otherwise, along the second sampling direction, it will be connected with the current sampling node x new The nearest node x closestinB Mobile node limit distance S limit Go to a new node and update the new node to the current sampling node x new , then enter S2.6; S2.6: Use the three-dimensional model of the welding workpiece to analyze the current sampling node x new Perform collision detection. If the collision detection passes, the current sampling node x new Incorporate into the robot path graph B, otherwise regenerate the sampling node x new , sampling node x new There is a probability P of a random node in the robot path graph A until the current sampling node x new The distance dis(x new ,x closestinB ) is less than or equal to the node limit distance S lim And through collision detection, it is put into the robot path graph B, and then in the robot path graph B, the current sampling node x is calculated according to the path cost new Perform partial graph reorganization to obtain the node order of the robot path graph B; S2.7: Repeat S2.1 to S2.6 to obtain the node order in the final robot path graph A and robot path graph B, and then select the node order from the current sampling node x. new After starting backtracking, the initial planned path is found.

5. The method for robot flexible welding path planning for non-standard workpieces according to claim 4, characterized in that: The current sampling node x is calculated based on the path cost. new Partial graph reorganization is performed, specifically: In robot path graph A or robot path graph B, for the current sampling node x new Each node x in the neighborhood N neighbo , calculate through the current sampling node x new Connect to neighboring node x neighbor If the new path cost is less than the neighboring node x neighbor The current path cost is , then replace the current sampling node x new is the neighboring node x neighbor The parent node; then calculate through the neighboring node x neighbor Connect to the current sampling node x new If the new path cost is less than the current sampling node x new The current path cost of the node is replaced by the neighboring node x neighbor is the current sampling node x new The parent node of .

6. A robot flexible welding path planning method for non-standard workpieces according to claim 4, characterized in that: The formula of the path cost is as follows: C=ω1C length +ω2C collision +ω3C joint_limits +ω4C stability +ω5C smoothness Among them, C represents the path cost, C length represents the path length, C collision represents the collision risk, C joint_limits represents the robot joint limit, C stability represents the robot posture stability, C smoothness represents motion smoothness; ω1~ω5 represent the first to fifth weights.

7. The method for robot flexible welding path planning for non-standard workpieces according to claim 1, characterized in that: The S3 is specifically: S3.1: Construct the measurement matrix M based on the initial planning path. After performing singular value decomposition and calculation on the measurement matrix M, the super ellipsoid rotation matrix C is obtained. The formula is as follows: U∑V T ≡M C=Udiag{1,...,det(U)det(V)}V T Among them, a1 represents the starting point x of the initial planning path start To the end point x of the initial planned path end direction, U,∑,V T are the left singular vector matrix, diagonal matrix and right singular vector transposed matrix respectively, ||||2 represents the Euclidean norm; I1 is the 1*n dimensional identity matrix, T represents the transpose, det() represents the determinant of the matrix; diag{} represents the diagonal matrix, and {} represents the elements on the diagonal of the matrix; S3.2: Based on the length of the current planned path and the starting point x start and the end point x end To establish a super ellipsoid, the formulas for the relevant geometric parameters are as follows: Among them, a, b, and c represent the major semi-axis, minor semi-axis, and focal length of the super ellipsoid, respectively, and c best Represents the current planned path length, c min Represents the starting point x of the current planned path start and the end point x end The distance between S3.3: Construct the shape matrix L of the super ellipsoid, the formula is as follows: S3.4: Randomly sample a node x in the unit hypersphere ball , through the shape matrix L and the super ellipsoid rotation matrix C, the node x ball After scaling, rotating and translating, we get the sampling point x' inside the super ellipsoid new ; S3.5: Traverse all nodes in robot path graph A and robot path graph B and the current sampling point x' new The distance between the current sampling point x' new The closest node x closet The corresponding distance dis(x' new ,x closest ) is less than or equal to the node limit distance S limit , then enter S3.6; otherwise, along the third sampling direction, it will be connected to the current sampling point x' new The closest node x closet Mobile node limit distance S limit Go to a new node and update the new node to the sampling point x' new , then enter S3.6; S3.6: Use the three-dimensional model of the welding workpiece to calculate the current sampling point x' new Perform collision detection. If the collision detection passes, the current sampling point x' new Included in robot path diagram A or robot path B, otherwise execute S3.4 to S3.5 again until the current sampling point x' new Through collision detection and placed in robot path map A or robot path B; S3.7: Dynamically adjust the size of the neighborhood N based on the cost function, the formula is as follows: Among them, N size Represents the adjusted domain size, N D Represents the original domain size, L current Represents the current optimal path length, L baseline represents the reference path length, β represents the adjustable dynamic parameter that controls the rate; S3.8: Based on the current neighborhood size N, the current sampling node x' is sampled according to the path cost new Perform partial graph reorganization to obtain a new node order in the robot path graph A or the robot path graph B; S3.9: Repeat S3.2-S3.8 to obtain the optimal planning path.

Images