Robot control based on orientated object detector

• Main Author: Li, Yongjie
• Other Authors: Cheah Chien Chern
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/173322
• Institution: Nanyang Technological University

This dissertation introduces an innovative solution for robot arm control based on oriented object detection. The proposed solution integrates a rotated object detection algorithm into the robot control system. The system comprises the D435i depth camera for image capture and the UR5e robot arm for robot arm manipulation to demonstrate the feasibility of this solution. To enhance the system’s detection capabilities, a Convolutional Neural Network (CNN)-based YOLOv5 algorithm with oriented bounding boxes is employed for detecting rotated objects. This technological advancement improves the accuracy of identifying and localizing objects, which is crucial for effective robot arm control. The seamless integration of the YOLOv5 OBB algorithm into the robotic arm enables efficient and precise control. Through experimental validation, the system showcases its effectiveness in achieving a high degree of precision in the control process. The integration of rotated object detection and robotic technologies positions this combination as a valuable solution for enhancing efficiency in applications. This dissertation contributes to the evolving landscape of implementing YOLObased rotated object detection neural networks into robot control systems. The dissertation methodically employed this system to execute a watering task, thereby substantiating the practical viability of the proposed solution