Object Manipulation using Robotic Hands with Varying Degrees of Grasp Knowledge.

Abstract

To date, there is limited rigorous analysis of object manipulation using robotic hands, where more focus has been placed on heuristic and experiment-based approaches. In this paper, we analyze the effects of grasp uncertainties based on realistic assumptions and propose a robust control framework for object manipulation. The framework considers a hand-object system subject to disturbances resulting from uncertainties in the object center of mass/inertia, hand kinematics, external wrenches, and contact locations. The proposed framework is then applied to practical object manipulation scenarios with different levels of uncertainty related to the sensors available to the robotic hand. These scenarios include when the hand-object system is known perfectly; when vision sensors are available; when tactile sensors are available; and when no vision/tactile sensors are available to the robotic hand (i.e. blind grasping). The analysis also addresses the internal force control in relation to the various practical cases. Simulation and experimental results validate the effectiveness of the proposed approach.