Blind convex hull shape estimation using passive-unwinding of a belt-pulley system attached to a gripper

This paper proposes a novel method for estimating the convex hull shape of an object without visual information using a gripper equipped with a belt-pulley system at the fingertips. When the taut belt comes into contact with an object, the belt unwinds passively via the pulley. Leveraging the passive unwinding and winding motion of the belt, shape sensing is achieved effectively over a wide tensioned area with a simple mechanism and minimal information requirements. We discuss the developed gripper, passive unwinding and winding mechanisms, gripper inputs, and the belt angle estimation method. Then, a two-dimensional analysis model is introduced, and an estimation problem for the belt angles is introduced. Next, a convex hull shape estimation algorithm is proposed using the estimated belt angles. Finally, actual machine experiments are conducted using the developed gripper. The results show that the belt adaptively deforms to objects with various shapes via passive unwinding and winding, and the proposed estimation method for the convex hull shape is validated.