A 3D Point Cloud Registration Algorithm Generalized Demonstration Trajectory Scheme

TL;DR: In this article , a technique for detecting changes in the front and back locations of the workpiece using point cloud registration and transforming the original grinding trajectory, eliminating the need for re-teaching is presented.

Abstract: Industrial robots are becoming more significant as the age of intelligent production progresses. Grinding robots have substantially emancipated the labor force in the industrial business. To get the grinding trajectory, many grinding robots still employ the teaching method, which takes a long period. The initial grinding trajectory becomes invalid when the relative location of the grinding robot and the workpiece changes, and it must be taught anew. This research provides a technique for detecting changes in the front and back locations of the workpiece using point cloud registration and transforming the original grinding trajectory, eliminating the need for re-teaching. Single-point instruction and dynamic movement primitives (DMP) generalization are used to create the reference trajectory. The sample consistency initial alignment (SAC-IA) algorithm-based improved matching approach and the Iterative closest point (ICP) algorithm collects point cloud data before and after the workpiece, and the transformation matrix of the front and rear locations of the workpiece is generated. It is used to change the original grinding trajectory, allowing the grinding trajectory to be reused. The suggested approach outperforms the ICP and SAC-IA+NDT algorithms in terms of resilience and accuracy, and saves a significant amount of time when compared to re-teaching.