With the widely used for the superplastic forming/diffusion bonding, the ladder height difference defects (LHDDs) affect the quality of the products. The uncertain nature of LHDDs imposes a challenge for the robot to accurately detect and plan the grinding path. Based on the above problems, the purpose of this study is to propose an online path planning method for LHDDs grinding using three-dimensional (3D) point cloud.
The robotic grinding system is composed of an industrial robot system, a force-controlled grinding device of machine tool, an industrial-strength 3D camera and a PC. The robotic grinding system detects and identifies LHDDs with 3D point cloud, plans the grinding path, then controls the robotic arm for grinding.
The experimental results show that the proposed method can identify LHDDs path and plan robotic grinding path automatically. The smooth robotic grinding path is planned based on the position and pose.
A LHDD path enhancement algorithm based on gradient variance feature is proposed to accurately extract feature points of LHDD paths in depth images. Pose estimation algorithm of the grinding points based on the adjacent point clouds of the LHDD path is proposed to solves the interference on the normal vector estimation of grinding points from LHDDs.
