A robotic manipulator demonstrates higher consistency than human in tasks that require precision and repeatability. As a result, the automated robotic manipulators are widely adopted in fields such as factory and surgery.
However, these manipulators are operated only in the controlled environments, following pre-programmed instructions. Current robots accompany re-programming or re-teaching when encountering unseen environments/objects. They are also limited in performing majority of human tasks, such as knot-tying, rope untying, cable harnesses, and part assemblies.
Humans, in contrast, exhibit excellent flexibility in handling unseen objects under new environments. We are skilled to manipulate deformable objects on a daily basis. Our goal is to develop robotic technology capable of object manipulation with human-level agility, precision, and intelligence.
To achieve these goals, we are currently focusing on the following issues:
- Real-time visual servoing control for dynamic object manipulation
- Robust object/robot pose estimation under occlusion
- Automating knot-tying with 3D rope reconstruction
- Optimized grasping for handover in bimanual assembly tasks
- Automating precise assembly tasks in confined spaces
- Efficient robot/hand-eye calibration method
- Stable object manipulation based on visual and tactile features under occlusion
- Design and control of wearable extendable robot arm