Human-Arm Roll Estimation in Underactuated Grippers with Proprioceptive Feedback

Abstract

A method for the estimation of the roll angle of a human forearm, when grasped by a robot with an underactuated gripper, using proprioceptive information only, is being developed. Knowing the angle around the forearm's axis (i.e. roll angle) is key for the safe manipulation of the human limb and biomedical sensor placement among others. The adaptive gripper has two independent underactuated fingers with two phalanges and a single actuator each. The final joint position of the gripper provides information related to the shape of the grasped object without the need for external contact or force sensors. Regression methods to estimate the roll angle of the grasping have been trained with forearm grasping information from different humans at each angular position. The results show that it is possible to accurately estimate the rolling angle of the human arm, not only for known people but also for humans for which haven't been previously trained.