Myoelectric teleoperation of a complex robotic hand

K.A. Farry, Ian D. Walker, Richard G. Baraniuk

IEEE Transactions on Robotics and Automation · 1996 · 249 citations · 33 references

Concepts

TL;DR

Teleoperation is a primary control mode for robots with complex hands, and myoelectric interfaces have been used for prosthetic hands for decades, but their feasibility for commanding multi‑degree‑of‑freedom grasps remains uncertain. This study presents a novel teleoperation method that converts the operator’s myoelectric signals into commands that replicate the motion of a complex anthropomorphic robotic hand. The method is implemented on NASA/Johnson Space Center’s 16‑DOF Utah/MIT Dextrous Hand, enabling two grasp types (key and chuck) and three thumb motions (abduction, extension, flexion) through signal‑processing, data‑collection apparatus, and a realtime teleoperation system. Realtime experiments achieved a 90 % correct grasp selection rate and an 87 % correct thumb motion selection rate using the myoelectric spectrum.

Abstract

Teleoperation continues to be a primary control mode in robotics applications, particularly for robots with complex hands. This paper details a novel method of teleoperation of complex anthropomorphic robotic hands: converting the myoelectric signal (generated by the operator's muscles during movement) into robot commands replicating the motion. Myoelectric prosthetic hands have used this user interface for over two decades; however, the feasibility of using this approach for commanding more than one degree-of-freedom, as in the pincher type grip in current myoelectric hands, has been in question. The research described in this paper addresses myoelectric control of NASA/Johnson Space Center's sixteen degree-of-freedom Utah/MIT Dextrous Hand for two grasping (key and chuck) options and three thumb motions (abduction, extension, and flexion). We discuss myoelectric signal processing approaches, data collection apparatus, and a realtime teleoperation implementation. We also present results in realtime discrimination of key and chuck grasps and offline discrimination of thumb motions. Our results include a 90% correct grasp selection rate and an 87% correct thumb motion selection, both using the myoelectric spectrum.

References

33