Thoranin Oonariya.. Implementing advanced object interaction in rescue robotics using an autonomous manipulator arm. Master's Degree(Automation Engineering). King Mongkut's University of Technology North Bangkok. Central Library. : King Mongkut's University of Technology North Bangkok, 2024.
Implementing advanced object interaction in rescue robotics using an autonomous manipulator arm
Abstract:
Manipulator arms in rescue robots remain a challenge in both competitions and real-life situations. In this paper, we present an autonomous in-house manipulator system specifically engineered to enrich object interaction in rescue robotics for the RoboCup Rescue Robot Competition 2024 as a contestant under the iRAP Robot Team. The manipulator is equipped with an RGB-D camera, providing pose estimation of objects by leveraging 3D models generated through prior scanning. YOLOv8 2D object detection is employed to identify objects in the scene, and cropped depth images are used to subsequently generate object point clouds. Iterative Closest Point (ICP) techniques, including point-to-plane, point-to-point, and color registration, are utilized in conjunction to provide accurate alignment between the object's point cloud and the pre-scanned 3D models. These techniques work together to minimize Root Mean Square Error (RMSE) during the alignment process, ensuring reliability of pose estimation before executing robotic movement actions. The system also uses IMU data from the RGB-D camera for manipulator impact detection. The entire system is developed within the Robot Operating System (ROS) framework, utilizing MoveIt for motion planning and control. The methodologies and innovations introduced in this work are adapted for deployment in the challenging and unexpected environments of the RoboCup Rescue Robot Competition.