Object pose estimation using RGB-D camera and robotic manipulator

Abstract

This work proposes a ROS2-based pipeline for object pose estimation, using a robotic manipulator and an RGB-D camera. The proposed pipeline includes a set of coordinated nodes that address key sub-problems: camera intrinsic calibration, hand-eye calibration, point cloud stitching, object detection, and object pose estimation. A detailed and comprehensive description of the solution design and implementation is provided, along with insights into common issues encountered during the testing phase. Finally, experimental results are presented to evaluate the system's effectiveness, and potential improvements are discussed.