Autonomous Vision-based Control for Within-hand Manipulation using Variable Friction Fingers

Caleb Chanthi Raj, Joshua Amrith Raj

Etd

Autonomous Vision-based Control for Within-hand Manipulation using Variable Friction Fingers

Public

Robotic within-hand manipulation is challenging to implement, even for highly articulated, sensorized and expensive robotic hands, mainly due to the lack of accurate hand-object and contact models. In this thesis, automatic manipulation planning and control schemes are investigated for a robotic hand with variable friction finger surfaces. Three strategies, 1) an offline motion planner that is utilized in a feedforward manner, 2) an online vision-based control algorithm, 3) a hybrid approach that combines the offline and online algorithms, are proposed to reposition and reorient an object within the workspace of the hand. These methods are designed considering the non-holonomic and switching nature of the system. The experimental results show that each method provides different advantages in terms of efficiency, accuracy, and path trajectory smoothness. A method to track the object without the presence of any marker is also presented.

Creator