Research On Path Planning Of 6 Degrees Of Freedom Manipulator Arm Based On Machine Vision

With the continuous progress and development of contemporary science and technology,industrial robotic arms have been widely used in various fields.Nowadays,in the face of increasingly complex operating environments,higher requirements are put forward for the efficiency,precision and stability of robotic arms to perform tasks.This paper takes ABB IRB-120 manipulator as the object to study the path planning problem of manipulator based on machine vision,and the main contents are as follows:First,a vision-based robotic arm experimental system was built.According to the needs of the robotic arm combined with machine vision to complete the autonomous path planning,the robotic arm experimental platform was built through the hardware of the robotic arm body,control cabinet,depth camera and ROS system software,and the creation of the robotic arm model under the ROS system was completed,and the calibration experiment between the robotic arm and the depth camera was completed,and the conversion relationship between the base coordinate system of the robotic arm and the camera coordinate system was obtained.The D-H parameter method was used to establish the kinematics model of the manipulator.Second,an object detection algorithm based on the improved YOLOv5 s network structure is given,which effectively improves the responsiveness of the system.Aiming at the problem that the system response ability is not strong due to the large amount of computation of YOLOv5 s network structure,the Mobile Net V3-YOLOv5 s lightweight network is obtained by replacing CSPDarknet as the backbone feature extraction network of the algorithm by replacing the backbone feature extraction network of the lightweight network Mobile Net V3,and the performance of the two types of network models is compared through the training and detection of self-made datasets,and the size of the improved network model is about 50%.Approximately 27% faster inspection.The test results show that the improved Mobile Net V3-YOLOv5 s algorithm not only reduces the amount of network calculation,optimizes the size of the algorithm model,improves the response performance of the whole system,but also ensures the accurate recognition and positioning of objects by the robotic arm.Third,an improved RRT-connect algorithm is proposed,which improves the path planning speed of the robotic arm in the obstacle environment and shortens the path length.Aiming at the problem that the RRT-connect algorithm has strong randomness and fixed step size in the expansion mode,the RRT-connect algorithm is improved by introducing a dynamic step size to improve the RRT-connect algorithm by introducing a dynamic step size to improve the RRT-connect algorithm by taking the middle position of the starting point and the target point as each random sampling point.Through simulation experiments,the results show that the planning time of the improved algorithm is reduced by about 28%,and the planned path length is shortened by about 22%,which proves that the proposed algorithm is effective.The two algorithms are applied to the robotic arm experimental platform for verification,and the improved algorithm planning time is reduced by about 27%,and the manipulator operation time is reduced by about 23%,which is basically consistent with the simulation results,which proves the applicability and superiority of the improved algorithm applied on the robotic arm experimental platform.