Dexterous in-hand manipulation tasks have been difficult to execute, even with highly complex hands and control schemes, as the object grasp stability needs to be maintained while it is displaced in the hand workspace. Researchers have shown that underactuated, adaptive hand designs can effectively immobilize objects with simple, open-loop, but there have been few cases where underactuation has been leveraged to enhance in-hand manipulation. In this work, we investigate the performance of a gripper utilizing a thumb with an active, belt-driven, conveyor surface and an opposing, underactuated finger with passive rollers, for a variety of manipulation tasks and range of objects. We show that consistent, repeatable object motion can be obtained while ensuring a rigid grasp without a priori knowledge of the object geometry or contact locations, due to the adaptive qualities of underactuated design. Many dexterous in-hand manipulation examples with their anthropomorphic equivalents are examined, and simple, open-loop control schemes to optimize the repeatability of these tasks are proposed.
展开▼