Project Role and Responsibilities

My work focuses on vision-based robot control and remote robotic manipulation for hazardous environments, particularly for applications related to nuclear facilities and non-destructive inspection.

My main responsibilities include:

• Developing vision-based robot control systems for remote inspection and manipulation
• Integrating VR-based teleoperation with autonomous visual servoing
• Developing object recognition, pose estimation, and target-tracking methods using RGB-D cameras
• Implementing robot–camera calibration and coordinate transformations for precise manipulation
• Integrating Unity, ROS2, robot manipulators, and vision sensors for real-time control
• Designing robot scanning and manipulation strategies for inspection tasks in environments where direct human access is limited

Project Contributions

Vision-Based Robot Control and System Integration for Non-Destructive Inspection
Developed and integrated a vision-based robot control system for non-destructive inspection in environments where direct human access is limited. Integrated an xArm7 manipulator with Intel RealSense cameras in a ROS2-based environment and performed eye-in-hand calibration using April Tag-based tracking and hand–eye calibration methods. Established coordinate transformations between the camera and robot frames to convert visual measurements into robot motion commands, enabling real-time vision-guided inspection and manipulation.

Deep Learning-Based Object Recognition and Pose Estimation
Developed deep learning-based vision methods for detecting and segmenting industrial objects such as pipes using YOLO-based instance segmentation. Processed RGB-D data to estimate object position, orientation, center axis, and candidate grasping points for robotic manipulation. The estimated geometric information was further used to support robot alignment, target tracking, and inspection-path generation.