LAUSR

Robust object manipulation using a robotic hand remains a challenging task, especially when a priori knowledge of the object being manipulated is not available. Tactile-based blind grasping is hereby defined where only typical sensors on-board the robotic hand unit are available. In this letter, we propose a robust discrete-time controller for tactile-based blind grasping to manipulate unknown objects. The analysis contained herein shows that from an initial stable pose, tactile feedback in the form of contact location and joint measurements are sufficient information to perform object manipulation, and ensure that the object does not slip. Furthermore, the analysis guarantees semiglobal practical asymptotic stability of the closed-loop system and takes into consideration the effect of sampling time. The robustness of the proposed control is demonstrated through simulation and experiments.