A collision-free path planning method for welding robots

Abstract

The application discloses a kind of welding robot collision-free path planning methods, belong to industrial robot motion planning technical field.Method includes: in the expansion process of search tree, with current expansion node as starting point, calculate local obstacle to current expansion node's improved repulsion;Improved repulsion includes the main repulsion component along the direction of obstacle pointing to current expansion node and the propulsion component along the direction of current expansion node pointing to target node;Improved repulsion is superimposed with target node attractive force, random sampling node attractive force, to obtain artificial potential field combined force direction;According to local obstacle risk value, determine the fusion weight of reference attractive direction and artificial potential field combined force direction, to calculate fusion expansion direction;The direction consistency index of fusion expansion direction and reference attractive direction is calculated, and based on direction consistency index and allowable deflection angle determine final expansion direction, to complete search tree.The application can realize efficient, stable, high-precision collision-free path planning.

Description

Technical Field

[0001] This invention relates to the field of industrial robot motion planning technology, and in particular to a collision-free path planning method for welding robots. Background Technology

[0002] Structural component welding is a crucial process in the manufacturing of engineering machinery, ships, steel structures, rail transportation, and aerospace equipment. As the manufacturing industry moves towards flexibility, digitalization, and intelligence, welding robots are increasingly being used in automated welding operations for various structural components due to their advantages such as high motion repeatability, good operational stability, high production efficiency, and reduced manual labor intensity. For diverse, small-batch, and non-standard structural components, there are often significant differences in the geometric dimensions, weld locations, clamping postures, and surrounding working environments of different workpieces.

[0003] Traditional robot teaching and programming methods require operators to adjust the welding torch's trajectory point-by-point according to the specific structural component. When the component model, installation location, or weld position changes, re-teaching and program debugging are usually necessary, resulting in a long programming cycle and making it difficult to meet the needs of flexible welding production. Because structural components typically have complex shapes, narrow local spaces, and irregular obstacle distributions, the collision-free paths available for robot movement between adjacent welds are limited. Especially inside structural components, in corner areas, or in areas with dense obstacles, the feasible configuration distribution in the robot's joint space is highly nonlinear and discontinuous. Randomly generated robot configurations and connection paths are prone to collisions with structural components or the surrounding environment.

[0004] Therefore, it is necessary to provide an efficient collision-free path planning method for structural component welding robots, which can quickly plan the motion paths of each joint of the robot, thereby improving the success rate of path planning and motion efficiency while avoiding collisions. Summary of the Invention

[0005] This invention provides a collision-free path planning method for welding robots. The technical solution is as follows: On the one hand, a collision-free path planning method for a welding robot is provided, the method comprising: Based on the robot kinematics model, the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space are obtained. The first configuration is used as the starting node and the second configuration is used as the target node to construct a search tree. During the expansion of the search tree, starting from the current expansion node, the improved repulsive force of local obstacles on the current expansion node is calculated; the improved repulsive force includes the main repulsive force component along the direction from the obstacle to the current expansion node, and the propulsive component along the direction from the current expansion node to the target node. The improved repulsive force is superimposed with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force. Based on the local obstacle distribution, calculate the first local obstacle risk value of the current expansion node along the reference attraction direction; the reference attraction direction is determined by combining the attraction force of the target node and the attraction force of the randomly sampled node. Based on the first local obstacle risk value, determine the fusion weight of the reference attraction direction and the artificial potential field force direction, and calculate the fusion expansion direction based on the fusion weight; Calculate the directional consistency index between the fusion extension direction and the reference attraction direction, and determine the final extension direction based on the directional consistency index and the allowable deflection angle; Along the final expansion direction, expand the step size to generate candidate new nodes; determine whether the candidate new node collides with an obstacle, and if not, add the candidate new node to the search tree.

[0006] On the other hand, a collision-free path planning device for a welding robot is provided, the device comprising: The acquisition unit is used to acquire the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space based on the robot kinematics model, so as to construct a search tree with the first configuration as the starting node and the second configuration as the target node. The first calculation unit is used to calculate, starting from the current expansion node, an improved repulsive force of local obstacles on the current expansion node during the expansion of the search tree; the improved repulsive force includes a main repulsive force component along the direction from the obstacle to the current expansion node, and a propulsive component along the direction from the current expansion node to the target node. The superposition unit is used to superimpose the improved repulsive force with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force. The second calculation unit is used to calculate the local obstacle risk value of the current extended node along the reference attraction direction based on the first local obstacle distribution; the reference attraction direction is determined by combining the attraction force of the target node and the attraction force of the randomly sampled node. The first determining unit is configured to determine the fusion weight of the reference attraction direction and the artificial potential field force direction based on the first local obstacle risk value, and calculate the fusion expansion direction based on the fusion weight; The second determining unit is used to calculate the directional consistency index between the fusion extension direction and the reference attraction direction, and to determine the final extension direction based on the directional consistency index and the allowable deflection angle. The generation unit is used to generate candidate new nodes by expanding the step size along the final expansion direction; to determine whether the candidate new node collides with an obstacle, and if not, to add the candidate new node to the search tree.

[0007] On the other hand, a computer device is provided, the computer device including a memory and a processor, the memory for storing computer programs, and the processor for executing the computer programs stored in the memory to implement the steps of the above-described collision-free path planning method for welding robots.

[0008] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, the steps of the above-described collision-free path planning method for welding robots are implemented.

[0009] On the other hand, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the above-described collision-free path planning method for welding robots.

[0010] The technical solution provided by this invention can bring at least the following beneficial effects: By constructing an improved repulsive force that includes a principal repulsive force component along the direction from the obstacle to the current expansion node and a propulsive component along the direction from the current node to the target node, the interference of obstacles near the target node on the convergence of the search tree can be effectively reduced. Then, the target attraction, randomly sampled attraction, and improved repulsive force are superimposed to form an artificial potential force direction. The first local obstacle risk value along the reference attraction direction is calculated based on the local obstacle distribution. The fusion weight of the reference attraction direction and the artificial potential force direction is dynamically adjusted accordingly to generate a fused expansion direction. Next, the directional consistency index between the fused expansion direction and the reference attraction direction is calculated, and the final expansion direction is determined using the directional consistency index and the allowable deflection angle to complete the expansion of the search tree. This scheme systematically solves the problems of convergence difficulties, excessive deflection of the expansion direction, and difficulty in balancing efficiency and accuracy in complex welding scenarios caused by interference from obstacles near the target node, achieving efficient, stable, and high-precision collision-free path planning. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1This is a flowchart of a collision-free path planning method for a welding robot provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of an actual welding scenario provided by an embodiment of the present invention; Figure 3 This is a robot welding workpiece weld pose provided in one embodiment of the present invention; Figure 4 This is a schematic diagram of the MDH parameters of a robotic arm provided in an embodiment of the present invention; Figure 5 This is an improved artificial potential field model provided in one embodiment of the present invention; Figure 6 This is a direction-corrected artificial potential field model provided in one embodiment of the present invention; Figure 7 This is a schematic diagram of the actual welding effect of the robotic arm provided in an embodiment of the present invention; Figure 8 This is a structural diagram of a collision-free path planning device for a welding robot provided in an embodiment of the present invention; Figure 9 This is a hardware architecture diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0014] Analysis of existing path planning methods reveals the following drawbacks: Existing fast expanding random tree path planning methods mainly rely on random sampling to generate search nodes. In welding scenarios with complex structural parts and narrow local spaces, a large number of invalid nodes and collision connection edges are easily generated near obstacles, resulting in long path planning time and poor stability of planning results.

[0015] Existing fast-expanding random tree methods that integrate artificial potential fields typically superimpose the attractive forces of the target point, the attractive forces of random sampling points, and the repulsive forces of obstacles, and expand the search tree according to the direction of the resultant force. When the target location is near structural components or other obstacles, the repulsive force generated by the obstacles may interfere with the convergence of the search tree to the target location; when the repulsive force is large, it may also cause the expansion direction to deviate excessively from the target direction, resulting in the search tree detouring around obstacles.

[0016] Existing path planning methods typically employ a fixed step size or adjust the step size only based on the distance between the robot and obstacles, without fully considering the positional relationship of obstacles relative to the robot's actual direction of motion. In open areas, smaller step sizes reduce search efficiency; in areas with dense obstacles, larger step sizes are prone to collisions, making it difficult to balance path planning speed and local obstacle avoidance accuracy.

[0017] To address the aforementioned shortcomings, the inventive concept of this invention is as follows: In the application scenario of automated welding of structural components, based on the ending pose of the current weld, the starting pose of the next weld, and the structural component and its surrounding environment, the robot automatically plans a collision-free jump path between adjacent welds. This ensures that there are no collisions between the robotic arm links, welding torch, structural component to be welded, fixture, and surrounding equipment, while improving the success rate, search efficiency, and path quality of path planning. An artificial potential field guides the rapid expansion of a random tree for robot joint space path search. By adjusting the repulsive force of obstacles, the interference of obstacles near the target position on the convergence of the search tree to the target node is reduced. Based on the relationship between the target attraction direction and the force direction of the artificial potential field, the expansion direction of the search tree is adaptively corrected to avoid excessive deviation from the target direction. Simultaneously, the expansion step size is dynamically adjusted based on the obstacle distance and the positional relationship of the obstacle relative to the actual expansion direction, enabling the search tree to expand rapidly in open areas.

[0018] The following describes the specific implementation of the above concept.

[0019] Please refer to Figure 1 The present invention provides a collision-free path planning method for a welding robot, the method comprising: Step 100: Based on the robot kinematics model, obtain the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space, so as to construct a search tree with the first configuration as the starting node and the second configuration as the target node. Step 102: During the expansion of the search tree, starting from the current expansion node, calculate the improved repulsive force of the local obstacle on the current expansion node; the improved repulsive force includes the main repulsive force component along the direction from the obstacle to the current expansion node, and the propulsion component along the direction from the current expansion node to the target node. Step 104: The improved repulsive force is superimposed with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force. Step 106: Calculate the first local obstacle risk value of the current expanding node along the reference attraction direction based on the local obstacle distribution; the reference attraction direction is determined by combining the attraction force of the target node and the attraction force of the randomly sampled node. Step 108: Based on the first local obstacle risk value, determine the fusion weight of the reference attraction direction and the artificial potential field force direction, and calculate the fusion expansion direction based on the fusion weight; Step 110: Calculate the directional consistency index between the fusion extension direction and the reference attraction direction, and determine the final extension direction based on the directional consistency index and the allowable deflection angle; Step 112: Generate candidate new nodes by expanding the step size along the final expansion direction; determine whether the candidate new node collides with an obstacle. If it does not collide, add the candidate new node to the search tree.

[0020] In this embodiment of the invention, an improved repulsive force is constructed, comprising a main repulsive force component along the direction from the obstacle to the current expansion node and a propulsive component along the direction from the current node to the target node. This effectively reduces the interference of obstacles near the target node on the convergence of the search tree. The target attraction, randomly sampled attraction, and improved repulsive force are then superimposed to form an artificial potential force direction. A first local obstacle risk value along the reference attraction direction is calculated based on the local obstacle distribution. The fusion weight of the reference attraction direction and the artificial potential force direction is dynamically adjusted accordingly to generate a fused expansion direction. Then, the directional consistency index between the fused expansion direction and the reference attraction direction is calculated, and the final expansion direction is determined using the directional consistency index and the allowable deflection angle to complete the expansion of the search tree. This solution systematically solves the problems of convergence difficulties, excessive deflection of the expansion direction, and difficulty in balancing efficiency and accuracy in step size caused by interference from obstacles near the target node in complex welding scenarios, achieving efficient, stable, and high-precision collision-free path planning.

[0021] The following study investigates the collision-free jump path between adjacent weld seams when the Elite CS625 collaborative robotic arm equipped with a welding torch performs automated welding on structural components. A real-world welding scenario is as follows: Figure 2 As shown, the robot welding workpiece weld seam pose is as follows: Figure 3 As shown, when welding structural components, the poses of adjacent weld seams differ significantly, and the path search speed directly determines the efficiency of completing the welding task. Therefore, it is necessary to establish an effective path planning method to improve the search efficiency of collision-free weld seams. The following section uses this research as a case study to describe this process. Figure 1 The execution method for each step is shown.

[0022] First, for step 100, based on the robot kinematics model, the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space are obtained, so as to construct a search tree with the first configuration as the starting node and the second configuration as the target node.

[0023] In one embodiment of the present invention, the robot can be a six-degree-of-freedom serial robot, with a welding torch mounted at the end effector. Let the joint variables of the robot be... ,in, Indicates the first The joint angles of each joint. In this embodiment of the invention, the improved Denavit-Hartenberg parametric method can be used to establish the robot's kinematic model; please refer to [link to relevant documentation]. Figure 4 Based on the link length, link twist angle, joint offset, and joint angle of each link in the robot, the homogeneous transformation matrix between adjacent link coordinate systems is obtained. Multiplying the transformation matrices between adjacent coordinate systems sequentially yields the forward kinematic relationship from the robot's base coordinate system to the end-effector welding torch coordinate system, expressed by the following first formula: in, Let be the homogeneous transformation matrix from the robot's base coordinate system to the TCP coordinate system of the welding torch tool center point. For the robot The coordinate system of the first link and the first Homogeneous transformation matrix between the coordinate systems of the connecting rods This is the homogeneous transformation matrix from the robot's 6th coordinate system to the welding torch's TCP coordinate system.

[0024] Using the above forward kinematics relationship as the robot's kinematics model, the inverse kinematics can be solved by numerical iteration to obtain the first configuration of the current weld end pose in joint space, and the second configuration of the next weld start pose in joint space.

[0025] After establishing the kinematic model of the welding robot, this embodiment of the invention can use a fast expanding random tree algorithm guided by an artificial potential field to construct a search tree with the first configuration as the starting node and the second configuration as the target node, so as to search for a collision-free path between the starting node and the target node in the robot joint space.

[0026] Then, for step 102, during the expansion of the search tree, starting from the current expansion node, the improved repulsive force of the local obstacle on the current expansion node is calculated.

[0027] Traditional APF-RRT The Artificial Potential Field Rapidly-exploring Random TreeStar algorithm typically superimposes the attractive force of the target node, the attractive force of randomly sampled nodes, and the repulsive force of obstacles, and expands the search tree according to the direction of the artificial potential field force. However, as the search tree gradually approaches the target node and there are structural components, clamps, or other obstacles around the target node, the traditional obstacle repulsive force may be greater than the attractive force generated by the target node, causing the expansion direction of the search tree to deviate from the target node, and in severe cases, making it difficult for the search tree to enter the target area. To address this problem, this invention improves the obstacle repulsive force of the traditional artificial potential field, obtaining an improved repulsive force of local obstacles on the current expanding node; this improved repulsive force includes a main repulsive force component along the direction of the obstacle pointing to the current expanding node, and a propulsive component along the direction of the current expanding node pointing to the target node.

[0028] Specifically, let the current expansion node be... The target node is The distance from the current extended node to the target node is , No. The distance from each obstacle to the current node is The area affected by the obstacle is , No. The improved repulsive force of each obstacle on the current expanding node is calculated using the following second formula: in, For the first An improved repulsive force of an obstacle on the current expanding node; For the first An obstacle points in a unit direction towards the current expanding node. The unit direction from the current expanding node to the target node; The main repulsive force component along the obstacle pointing in the current direction of expansion. This is the propulsion component along the direction from the current extension node to the target node.

[0029] in, and They can be calculated using the following third and fourth formulas respectively: in, This is the repulsive force gain coefficient; The distance from the current expansion node to the target node. Power; For the current node to the th The square of the distance to each obstacle; This is the target distance adjustment factor.

[0030] This improved repulsive force introduces a target distance adjustment term into the traditional obstacle repulsive force. This causes the main repulsive force component along the direction from the obstacle to the current expanding node to gradually weaken as the distance between the current expanding node and the target node decreases, reducing the interference of obstacles near the target node on the convergence of the search tree. Simultaneously, a propulsion component along the direction from the current expanding node to the target node is added within the obstacle's influence range to maintain the trend of the search tree expanding towards the target region, thereby improving the search efficiency within the target node's neighborhood.

[0031] It should be noted that the starting node is the current expanding node at the time of the initial expansion, which is a temporary node being expanded in the search tree; in the initial state, the starting node is the current expanding node. During the iterative expansion process, the nearest node can be found by random sampling, and then expansion is performed. Collisions are checked based on the current expanding node. If there is no collision, it is added to the search tree to complete the optimization, until a new expanding node approaches or connects to the target node.

[0032] Next, explanations will be given for steps 104 ("superimposing the improved repulsive force with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the artificial potential force"), 106 ("calculating the first local obstacle risk value of the current expanding node along the reference attraction direction based on the local obstacle distribution"), and 108 ("determining the fusion weight of the reference attraction direction and the artificial potential force direction based on the first local obstacle risk value, and calculating the fusion expansion direction based on the fusion weight").

[0033] In this embodiment of the invention, in addition to the attraction of the target node to the current expanding node, the embodiment also considers the guiding role of randomly sampled nodes in the search tree expansion process. Specifically: Let the attractiveness of the target node be... The attractiveness of random sampling nodes is The direction of the artificial potential force is obtained by superimposing the improved repulsive force with the attractive force of the target node and the attractive force of the randomly sampled nodes. The artificial potential force acting on the currently extended node can be calculated using the following fifth formula: in, The current extension node receives the artificial visual field force; It refers to a set of local obstacles that are within the influence range of an obstacle.

[0034] By simultaneously introducing the attraction of target nodes and the attraction of random sampling points, we can maintain the trend of the search tree expanding towards the target region while preserving the RRT (Recovery Time Limit). The algorithm's random search capability prevents the search tree from expanding entirely in a single direction, which would reduce global search efficiency. Please refer to the improved artificial potential field model. Figure 5 . Figure 5 The meanings of each parameter are as follows: As the starting node, The nearest node, For the target node, For random sampling nodes, For the improved first obstacle repulsive force, For the improved second obstacle repulsion force, The attraction that the target node generates for the nearest node. The attraction of a randomly sampled node to its nearest neighbor. The artificial potential force is obtained by superimposing the attraction of the target node, the attraction of the randomly sampled node, and the improved repulsive force of each obstacle.

[0035] In this embodiment of the invention, without considering the repulsive force of obstacles, the reference attraction direction of the search tree is jointly constituted by the attractive force of the target node and the attractive force of the random sampling points. This reference attraction direction is calculated using the following sixth formula: in, The reference attraction direction, which includes both the guiding role of the target node and the exploratory role of the randomly sampled nodes, is used to identify the expected expansion trend of the search tree when the influence of obstacles is not considered.

[0036] After introducing the improved repulsive force of the obstacle, the expansion direction of the artificial potential field force can be calculated using the following seventh formula: in, The direction of the artificial potential field force after the introduction of the improved repulsive force of the obstacle is described below as the "direction of the artificial potential field force".

[0037] When the repulsive force generated by a local obstacle is strong, the direction of the force in the artificial potential field is... Possibly relative to the reference attraction direction A significant deflection occurs. If directly adopted... Expanding the search tree may cause it to veer excessively around obstacles; if it is fully adopted... This will weaken the obstacle avoidance effect of the repulsive force of the obstacle.

[0038] Based on this, in this embodiment of the invention, the reference attraction direction and the artificial potential field force direction are adaptively fused according to the obstacle distribution near the current expansion node. Let the first local obstacle risk value of the current expansion node along the reference attraction direction be... The larger the value, the greater the degree to which the current expanding node is affected by obstacles when expanding along the reference attraction direction.

[0039] In one embodiment of the present invention, the local obstacle risk value of the current extended node along the reference attraction direction is calculated as follows: The risk value of the first local obstacle is calculated using the following formula: in, The first local obstacle risk value of the current extended node along the reference attraction direction; The area affected by the obstacle; For the current node and the first The distance between the obstacles; For reference, the direction of attraction and the first The angle between the directions of the obstacles It refers to the set of local obstacles within the influence range of an obstacle.

[0040] The directional fusion weight is determined based on the risk value of the first local obstacle. The fusion weight is determined by the following eighth formula: in, and These are the lower and upper limits of the fusion weight, respectively. This represents the first local obstacle risk value of the current extended node along the reference attraction direction.

[0041] Based on the fusion weights, the reference attraction direction and the artificial potential field force direction are fused to obtain the fusion extension direction. This fusion extension direction is determined by the following ninth formula: in, This is for the direction of integration and expansion.

[0042] When the risk of the first local obstacle is small, the fusion weight is small, and the fusion expansion direction is closer to the reference attraction direction, so that the search tree maintains a strong target advancement capability; when the risk of the first local obstacle is large, the fusion weight increases, and the fusion expansion direction is closer to the artificial potential field force direction, thereby enhancing the obstacle avoidance capability of the search tree in the obstacle neighborhood.

[0043] Finally, the following explanations are given regarding step 110, "calculate the directional consistency index between the fusion expansion direction and the reference attraction direction, and determine the final expansion direction based on the directional consistency index and the allowable deflection angle," and step 112, "generate candidate new nodes along the final expansion direction with an expansion step size; determine whether the candidate new node collides with an obstacle; if it does not collide, add the candidate new node to the search tree."

[0044] To prevent the fusion extension direction from being excessively deflected relative to the reference attraction direction, in this embodiment of the invention, an allowable deflection angle can be further set to constrain the fusion extension direction.

[0045] Define the directional consistency index between the fusion extension direction and the reference attraction direction as follows: in, As an indicator of directional consistency, To fuse the extended direction vector; Let the allowable deflection angle corresponding to the current local obstacle risk be . Therefore, the allowable deflection angle can be calculated using the following formula (number ten): in, To allow for deflection angle; , These are the lower and upper limits of the allowable deflection angle, respectively.

[0046] In one embodiment of the present invention, determining the final expansion direction based on the directional consistency index and the allowable deflection angle includes: If the directional consistency index is lower than the cosine value corresponding to the allowable deflection angle, the fusion expansion direction is projected onto a cone centered on the reference attraction direction and with the allowable deflection angle as half an angle to generate the final expansion direction. If the directional consistency index is not lower than the cosine value corresponding to the allowable deflection angle, then the fusion expansion direction will be taken as the final expansion direction.

[0047] That is: when the following conditions are met When the deviation between the fusion extension direction and the reference attraction direction is within the allowable range, the fusion extension direction is directly taken as the final extension direction. ; When satisfied At this time, the fusion extension direction deflects too much relative to the reference attraction direction. In one embodiment of the present invention, the lateral component of the fusion extension direction relative to the reference attraction direction can be extracted according to the following eleventh formula: in, This refers to the lateral component of the fusion extension direction relative to the reference attraction direction.

[0048] The final expansion direction was revised to: Through the aforementioned directional correction, the final expansion direction of the search tree is restricted to a direction centered on the reference attraction direction, within an allowable deflection angle. Within the range of half-angle directions, such as Figure 6 As shown, in a relatively open local environment, the search tree mainly expands along the reference direction jointly formed by the target and randomly sampled nodes; in areas with dense obstacles, the expansion direction is allowed to be appropriately deflected by the artificial potential field obstacle avoidance direction, thereby maintaining the target's advancement capability while preserving local obstacle avoidance capability, and reducing detour expansion and invalid nodes in the obstacle neighborhood.

[0049] Traditional APF-RRT A fixed expansion step size is typically used, which makes it difficult to simultaneously achieve search efficiency in open areas and obstacle avoidance accuracy in the obstacle neighborhood. In one embodiment of the present invention, to distinguish it from the first local obstacle risk value calculated along the reference attraction direction, a second local obstacle risk value is calculated based on the final expansion direction and the local obstacle distribution, and the expansion step size is dynamically determined based on the second local obstacle risk value.

[0050] Let the current expansion node be The final expansion direction is The current expansion node and the first The distance between the obstacles is The area affected by the obstacle is Ultimate expansion direction The angle between the direction of the obstacle and the direction of the obstacle is The risk of the second local obstacle is defined as follows (Formula XIII): in, This refers to the set of local obstacles within the area of ​​influence of the obstacle. When there is a nearby obstacle in front of the final direction of expansion, Increase; when the node is in an open area or the obstacle is behind the final expansion direction. Reduce. The adaptive expansion step size is determined based on the second local obstacle risk using the following fourteenth formula: in, To extend the step size; , These are the lower and upper limits of the expansion step size, respectively.

[0051] The generated candidate new nodes are: When the risk of local obstacles is low, the expansion step size is close to the maximum step size, allowing the search tree to expand rapidly in open areas; when there is a nearby obstacle in front of the final expansion direction, the expansion step size is reduced, allowing the search tree to expand finely in the neighborhood of the obstacle, thereby reducing the number of invalid nodes and collision expansions.

[0052] The actual welding effect of the robotic arm is as follows Figure 7 As shown.

[0053] In this embodiment of the invention, the convergence ability of the search tree to the target area is improved, and the probability of path planning failure when the target location is near obstacles is reduced; invalid expansion and redundant search nodes near obstacles are reduced, and the planning time of collision-free paths between adjacent welds is shortened; the stability and success rate of path planning in complex structural components, narrow spaces and irregular obstacle distribution scenarios are improved.

[0054] It should also be noted that, in addition to being applicable to collision-free jump path planning between weld seams of structural components, the embodiments of the present invention can also be extended to industrial application scenarios such as robot grinding, spraying, cutting, assembly and inspection, which require collision-free movement between multiple working poses, so as to improve the path planning efficiency and adaptability of robots in complex obstacle environments.

[0055] Please refer to Figure 8 This invention provides a collision-free path planning device for a welding robot, the device comprising: The acquisition unit is used to acquire the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space based on the robot kinematics model, so as to construct a search tree with the first configuration as the starting node and the second configuration as the target node. The first calculation unit is used to calculate, starting from the current expansion node, an improved repulsive force of local obstacles on the current expansion node during the expansion of the search tree; the improved repulsive force includes a main repulsive force component along the direction from the obstacle to the current expansion node, and a propulsive component along the direction from the current expansion node to the target node. The superposition unit is used to superimpose the improved repulsive force with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force. The second calculation unit is used to calculate the first local obstacle risk value of the current expansion node along the reference attraction direction based on the local obstacle distribution; the reference attraction direction is determined by combining the attraction force of the target node and the attraction force of the randomly sampled node. The first determining unit is configured to determine the fusion weight of the reference attraction direction and the artificial potential field force direction based on the first local obstacle risk value, and calculate the fusion expansion direction based on the fusion weight; The second determining unit is used to calculate the directional consistency index between the fusion extension direction and the reference attraction direction, and to determine the final extension direction based on the directional consistency index and the allowable deflection angle. The generation unit is used to generate candidate new nodes by expanding the step size along the final expansion direction; to determine whether the candidate new node collides with an obstacle, and if not, to add the candidate new node to the search tree.

[0056] In one embodiment of the present invention, calculating the improved repulsive force of the local obstacle on the current expanding node includes: The improved repulsive force of the i-th obstacle on the current expanding node is calculated using the following formula: in, For the first An improved repulsive force of an obstacle on the current expanding node; For the first An obstacle points in a unit direction towards the current expanding node. The unit direction from the current expanding node to the target node; The main repulsive force component along the obstacle pointing in the current direction of expansion. The propulsion component is the component along the direction from the current extended node to the target node; The main repulsive force component and the propulsive component are calculated using the following formulas: in, This is the repulsive force gain coefficient; The distance from the current expansion node to the target node. Power; For the current expansion node to the th The square of the distance to each obstacle; This is the target distance adjustment factor.

[0057] In one embodiment of the present invention, the improved repulsive force is superimposed with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force, including: The direction of the force in an artificial potential field is calculated using the following seventh formula: in, The direction of the force in the artificial potential field; To attract the target node, Attraction of randomly sampled nodes; For the i-th obstacle, there is an improved repulsive force on the current expanding node. It refers to a set of local obstacles that are within the influence range of an obstacle.

[0058] In one embodiment of the present invention, determining the fusion weight of the reference attraction direction and the artificial potential field force direction based on the first local obstacle risk value includes: The fusion weight is determined using the following formula. : in, and These are the lower and upper limits of the fusion weight, respectively. The first local obstacle risk value of the current extended node along the reference attraction direction; The calculation of the fusion extension direction based on the fusion weights includes: The fusion expansion direction is calculated using the following formula. : in, For reference, the direction of attraction, The direction of the force in an artificial potential field.

[0059] In one embodiment of the present invention, determining the final expansion direction based on the direction consistency index and the allowable deflection angle includes: If the directional consistency index is lower than the cosine value corresponding to the allowable deflection angle, the fusion expansion direction is projected onto a cone centered on the reference attraction direction and with the allowable deflection angle as half an angle to generate the final expansion direction. If the directional consistency index is not lower than the cosine value corresponding to the allowable deflection angle, then the fusion expansion direction will be taken as the final expansion direction.

[0060] In one embodiment of the present invention, the expansion step size is dynamically calculated based on a second local obstacle risk value, which is calculated based on the distance between the obstacle and the current expansion node and the directional relationship of the local obstacle relative to the final expansion direction.

[0061] It should be noted that the collision-free path planning device for welding robots provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the collision-free path planning device for welding robots provided in the above embodiments and the collision-free path planning method embodiments for welding robots belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0062] Embodiments of this application also provide a computer device, please refer to... Figure 9 The computer device includes a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, at least one program, code set or instruction set being loaded and executed by the processor to implement the collision-free path planning method for welding robots provided in the above-described method embodiments.

[0063] Embodiments of this application also provide a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the collision-free path planning method for welding robots provided in the above-described method embodiments.

[0064] Embodiments of this application also provide a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium and executes the computer program, causing the computer device to perform any of the collision-free path planning methods for welding robots described in the above embodiments.

[0065] For ease of description, the above systems or devices are described separately as various modules or units based on their functions. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware components.

[0066] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0067] Finally, it should be noted that in this document, relational terms such as first, second, third, and fourth are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0068] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A collision-free path planning method for a welding robot, characterized in that, The method includes: Based on the robot kinematics model, the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space are obtained. The first configuration is used as the starting node and the second configuration is used as the target node to construct a search tree. During the expansion of the search tree, starting from the current expansion node, the improved repulsive force of local obstacles on the current expansion node is calculated; the improved repulsive force includes the main repulsive force component along the direction from the obstacle to the current expansion node, and the propulsive component along the direction from the current expansion node to the target node. The improved repulsive force is superimposed with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force. Based on the local obstacle distribution, calculate the first local obstacle risk value of the current expansion node along the reference attraction direction; the reference attraction direction is determined by combining the attraction force of the target node and the attraction force of the randomly sampled node. Based on the first local obstacle risk value, determine the fusion weight of the reference attraction direction and the artificial potential field force direction, and calculate the fusion expansion direction based on the fusion weight; Calculate the directional consistency index between the fusion extension direction and the reference attraction direction, and determine the final extension direction based on the directional consistency index and the allowable deflection angle; Along the final expansion direction, expand the step size to generate candidate new nodes; determine whether the candidate new node collides with an obstacle, and if not, add the candidate new node to the search tree.

2. The method according to claim 1, characterized in that, The calculation of the improved repulsive force of local obstacles on the current expanding node includes: The improved repulsive force of the i-th obstacle on the current expanding node is calculated using the following formula: in, For the first An improved repulsive force of an obstacle on the current expanding node; For the first An obstacle points in a unit direction towards the current expanding node. The unit direction from the current expanding node to the target node; The main repulsive force component along the obstacle pointing in the current direction of expansion. The propulsion component is the component along the direction from the current extended node to the target node; The main repulsive force component and the propulsive component are calculated using the following formulas: in, This is the repulsive force gain coefficient; The distance from the current expansion node to the target node. Power; For the current expansion node to the th The square of the distance to each obstacle; This is the target distance adjustment factor.

3. The method according to claim 1, characterized in that, The improved repulsive force is superimposed with the attractive force of the target node and the attractive force of the randomly sampled nodes to obtain the direction of the artificial potential force, including: The direction of the force in an artificial potential field is calculated using the following seventh formula: in, The direction of the force in the artificial potential field; To attract the target node, Attraction of randomly sampled nodes; For the i-th obstacle, there is an improved repulsive force on the current expanding node. It refers to a set of local obstacles that are within the influence range of an obstacle.

4. The method according to claim 1, characterized in that, The step of determining the fusion weight between the reference attraction direction and the artificial potential field force direction based on the first local obstacle risk value includes: The fusion weight is determined using the following formula. : in, and These are the lower and upper limits of the fusion weight, respectively. The first local obstacle risk value of the current extended node along the reference attraction direction; The calculation of the fusion extension direction based on the fusion weights includes: The fusion expansion direction is calculated using the following formula. : in, For reference, the direction of attraction, The direction of the force in an artificial potential field.

5. The method according to claim 1, characterized in that, The determination of the final expansion direction based on the directional consistency index and the allowable deflection angle includes: If the directional consistency index is lower than the cosine value corresponding to the allowable deflection angle, the fusion expansion direction is projected onto a cone centered on the reference attraction direction and with the allowable deflection angle as half an angle to generate the final expansion direction. If the directional consistency index is not lower than the cosine value corresponding to the allowable deflection angle, then the fusion expansion direction will be taken as the final expansion direction.

6. The method according to any one of claims 1-5, characterized in that, The expansion step size is dynamically calculated based on the second local obstacle risk value, which is calculated according to the distance between the obstacle and the current expansion node and the directional relationship of the local obstacle relative to the final expansion direction.

7. A collision-free path planning device for a welding robot, characterized in that, The device includes: The acquisition unit is used to acquire the first configuration of the current weld end pose in the joint space and the second configuration of the next weld start pose in the joint space based on the robot kinematics model, so as to construct a search tree with the first configuration as the starting node and the second configuration as the target node. The first calculation unit is used to calculate, starting from the current expansion node, an improved repulsive force of local obstacles on the current expansion node during the expansion of the search tree; the improved repulsive force includes a main repulsive force component along the direction from the obstacle to the current expansion node, and a propulsive component along the direction from the current expansion node to the target node. The superposition unit is used to superimpose the improved repulsive force with the attractive force of the target node and the attractive force of the randomly sampled node to obtain the direction of the artificial potential force. The second calculation unit is used to calculate the first local obstacle risk value of the current expansion node along the reference attraction direction based on the local obstacle distribution; the reference attraction direction is determined by combining the attraction force of the target node and the attraction force of the randomly sampled node. The first determining unit is configured to determine the fusion weight of the reference attraction direction and the artificial potential field force direction based on the first local obstacle risk value, and calculate the fusion expansion direction based on the fusion weight; The second determining unit is used to calculate the directional consistency index between the fusion extension direction and the reference attraction direction, and to determine the final extension direction based on the directional consistency index and the allowable deflection angle. The generation unit is used to generate candidate new nodes by expanding the step size along the final expansion direction; to determine whether the candidate new node collides with an obstacle, and if not, to add the candidate new node to the search tree.

8. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1-6.

Images