Search

CN-121973249-A - Path motion control method and system for welding robot

CN121973249ACN 121973249 ACN121973249 ACN 121973249ACN-121973249-A

Abstract

The invention discloses a path motion control method and a system of a welding robot, which relate to the technical field of intelligent control of welding robots and comprise the steps of constructing an initial collision-free reference path; updating a residual strategy network and a current real-time position point cycle by cycle, correcting a residual path based on the updated residual strategy network and a current state set corresponding to the current real-time position point to obtain a residual corrected planning path, extracting a current local displacement section from the residual corrected planning path based on interpolation step length, and sequentially splicing the current local displacement sections corresponding to each control cycle to form a target walking path optimized based on an initial collision-free reference path. According to the method, through cycle-by-cycle self-adaptive correction and dynamic parameter adjustment, the communication limitation of fixed welding control parameters is broken, and the problems that in the prior art, the welding process effect is caused by difficulty in adapting to environmental changes in real time and poor self-adaptive transitional effect are solved.

Inventors

  • DENG LU
  • YAN YUNING
  • LIU WENYAO

Assignees

  • 湖南大学

Dates

Publication Date
20260505
Application Date
20260407

Claims (9)

  1. 1. A path motion control method of a welding robot, comprising the steps of: S10, constructing an initial collision-free reference path, wherein the initial collision-free reference path is a third-order Bezier curve and is provided with a path starting point and a path ending point, and the collision-free reference path is provided with a reference control point for restraining a bending shape and tangential directions; S20, updating a residual strategy network and a current real-time position point cycle by cycle, correcting a residual path based on the updated residual strategy network and a current state set corresponding to the current real-time position point to obtain a residual corrected planning path, and extracting a current local displacement section from the residual corrected planning path based on interpolation step length, wherein the method specifically comprises the following steps: During a control period t, constructing a residual path from a current real-time position to a path end point, and extracting a real-time path control point carried by the residual path, wherein t is more than or equal to 1, and t is a positive integer; obtaining an updated residual strategy network when the motion reward score of the control period t converges, wherein the motion reward score is used for representing obstacle avoidance safety, motion smoothness adaptation degree and motion jerk degree of a motion process; Based on the current state set, the updated residual strategy network prediction is utilized, a spatial position correction vector and a time scaling factor corresponding to the control period t are output, and the current state set comprises a distance vector between the current real-time position and the surface of the obstacle and a normal vector of the surface of the obstacle in the control period t; the space position correction vector is overlapped to the real-time path control point, and the residual path is corrected to obtain a residual correction planning path; calculating and acquiring an interpolation step length and a target drop point of the control period t based on the current real-time position, the time scaling factor and a preset speed, and updating the interpolation step length into a current local displacement section; executing the motion of the current local displacement segment to reach the target drop point, and eliminating the current local displacement segment from the residual correction planning path; Updating the target drop point to be the current real-time position and entering the next control period, and repeating the steps until the current real-time position moves to the position of the path end point; And S30, sequentially splicing the current local displacement segments corresponding to the control periods t to form a target walking path optimized based on the initial collision-free reference path.
  2. 2. The method for controlling path movement of a welding robot according to claim 1, wherein, In step S20, the method further includes the steps of: And combining output results of the residual strategy network updated by each control period t, and acquiring physical motion control parameters corresponding to each discrete time step based on a preset single period duration, wherein the physical motion control parameters comprise a motion speed parameter and/or a falling point direction vector of the target falling point.
  3. 3. The method for controlling path movement of a welding robot according to claim 2, wherein, The step S10 specifically includes: s11, obtaining geometric information of an obstacle and start and stop point information of a welding seam based on environmental perception; S12, constructing an axis alignment bounding box based on a preset safety margin and the geometric information of the obstacle; s13, generating an initial collision-free reference path based on a third-order Bezier curve by using the axis alignment bounding box and the weld seam start-stop point information, wherein the collision-free reference path is provided with the corresponding reference control points, and the reference control points comprise first control points for controlling the bending shape to correspond to the tangential direction And a second control point 。
  4. 4. The method for controlling path movement of a welding robot according to claim 2, wherein, Using the formula , Calculating and obtaining the sports prize score, wherein, Indicating that the athletic reward score is to be played, Represents a path space security evaluation item score for reflecting the position of the assumed drop point, Represents a motion time efficiency evaluation item score for reflecting the position of the assumed drop point, Representing the dynamic smooth evaluation term scores at the assumed drop point locations, Representing the pose alignment evaluation item score at the time of assuming the drop point position, Representing the arrival indication function, The end point of the path is indicated, Representing a drop point tolerance threshold value, 、 、 And Respectively representing a safety evaluation coefficient, a time efficiency evaluation coefficient, a dynamics smoothing evaluation coefficient and an ending point consistency rating coefficient, and updating the assumed falling point to the target falling point if the sports reward score converges.
  5. 5. The method for controlling path-motion of a welding robot according to claim 4, wherein, Using the formula Calculating and obtaining the path space safety evaluation item score, wherein, Representing the distance excitation weight(s), The collision penalty weight is indicated as such, Representing control periods The euclidean distance of the landing point location from the obstacle surface is assumed, In order to prevent the singular smoothing term of the denominator, Representing control periods And 1 if the assumed drop point position collides, and 0 if the assumed drop point position does not collide, 。
  6. 6. The method for controlling path-motion of a welding robot according to claim 4, wherein, Using the formula Calculating and acquiring the score of the exercise time efficiency evaluation item, Representing the current real-time location of the object, , Representing control periods Assuming an instantaneous resultant velocity corresponding to the location of the drop point, Indicating a preset speed.
  7. 7. The method for controlling path-motion of a welding robot according to claim 4, wherein, Using the formula Calculating and obtaining the score of the dynamic smooth evaluation item, Representing the assumed drop point position at time period t, 、 、 Respectively representing the current real-time position calculated in the first three periods of the time period t, Representing the preset single-period duration; Wherein if the first three periods do not exist in the time period t, then The value is 0.
  8. 8. The method for controlling path-motion of a welding robot according to claim 4, wherein, Using the formula Calculating and obtaining the attitude alignment evaluation item scores, Represents the end velocity vector assuming the location of the drop point, Representing a normal tangential direction vector of the second segment target weld at the start point.
  9. 9. A path-motion control system of a welding robot, characterized by comprising processing means for implementing the steps of the path-motion control method of a welding robot according to any one of claims 1 to 8.

Description

Path motion control method and system for welding robot Technical Field The invention relates to the technical field of intelligent control of welding robots, in particular to a path motion control method and system of a welding robot. Background Referring to fig. 1, in the welding of steel structures for ship manufacturing, bridge engineering and large-scale construction, in order to release the welding residual stress, a process over-welding hole is generally reserved at the joint of the stiffening rib and the main board, the over-welding hole usually exists in a fan shape, the orifice is smaller, the over-welding hole is used as a typical limited space narrow channel scene, the path planning difficulty is far higher than that of a general free space obstacle avoidance, and the welding robot is required to span the over-welding hole and precisely connect the next section of welding seam after the welding robot finishes one section of welding. In the configuration space of the robot, in a narrow canyon environment where two larger feasible regions are connected by only one narrow, elongated channel, a common random sampling algorithm (such as RRT) is difficult to sample to the effective point in the narrow canyon, resulting in planning failure or extreme distortion of the path. The existing path planning technology has the technical defects that when the existing path planning technology faces obstacles in narrow canyon environments, the existing path planning technology is low in idle stroke efficiency, the existing technology mostly adopts manual teaching or a portal-shaped (lifting-translating-falling) obstacle avoidance track based on fixed geometric rules, in order to ensure safety, an excessive safety margin is reserved, an invalid idle stroke path is overlong, the whole production takt is seriously dragged, motion impact and precision loss are caused, a curvature mutation exists at an inflection point of a traditional hard coding transition track, a mechanical arm joint is caused to bear larger Jerk (also called variable acceleration and Jerk) impact, the kinetic discontinuity not only aggravates equipment abrasion, but also can cause terminal residual vibration when a robot reaches a target point, the positioning precision and the arc guiding success rate of a next-stage welding seam starting point are directly influenced, the unstructured environment adaptability is poor, in practice, the over-welding precision, edge burrs and group pairing errors have high uncertainty, the strategy based on a traditional offline programming (OLP) or rigid obstacle avoidance algorithm (such as an artificial potential field method) is difficult to change in real-time environment, and interference between the edge and a workpiece is easy to occur. In summary, in the conventional welding robot during the obstacle avoidance control process, it is difficult to realize adaptive transition based on mechanical consideration, and in view of this, it is necessary to provide a path motion control method and system for the welding robot to solve or at least partially solve the above technical problems. Disclosure of Invention The path motion control method and system of the welding robot solve the technical problems that in the obstacle avoidance control process of the existing welding robot, self-adaptive transition is difficult to achieve, environment change is difficult to adapt in real time (welding control parameters are fixed), and the welding process effect is not ideal when collision interference exists between a welding gun and the edge of a workpiece. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: A path motion control method of a welding robot includes the steps of: S10, constructing an initial collision-free reference path, wherein the initial collision-free reference path is a third-order Bezier curve and is provided with a path starting point and a path ending point, and the collision-free reference path is provided with a reference control point for restraining a bending shape and tangential directions; S20, updating a residual strategy network and a current real-time position point cycle by cycle, correcting a residual path based on the updated residual strategy network and a current state set corresponding to the current real-time position point to obtain a residual corrected planning path, and extracting a current local displacement section from the residual corrected planning path based on interpolation step length, wherein the method specifically comprises the following steps: at control period t, a slave current real-time position is constructed Remaining path to path end pointExtracting a real-time path control point carried by the residual path P (t), wherein t is more than or equal to 1, and t is a positive integer; obtaining an updated residual strategy network when the motion reward score of the control period t converges, wherein the motion reward