Prosthetic arms have been done before, I’ve been thinking about them for years. I decided there was plenty of research in that field, and with Dean Kamen’s “Luke” prosthesis I thought there would be much faster advancements. As it turns out, there is still need for a cheap, dexterous robot arm that can more easily act like the human hand. Recently on Adafruit’s blog they posted this article, concerning a student who made his own robot arm from the shoulder down that could pretty accurately complete human gestures. I saw some robot arms that use a big row of servos for each finger, and each finger in the joint is not controllable; either every digit in the finger flexes or none of them do.
My thinking is that instead of using a row of clunky servos, what if there was a small linear screw-drive motor for each joint? Say three joints per finger, that’s 15 motors, plus a rotate motor for the thumb and pan-tilt head for the wrist, that’s 18 joints. And if I could find a sufficiently small motor, I could house all the joint mechanics within a shape much closer to the human forearm. The Screw-drive mechanism would allow much greater gripping strength, since threaded rods aren’t easily back driven. Also, little if any power would be required to keep a grip. This is unlike servos that face a rotational force, causing them to always exert force and maintain position. With the direction of travel at an angle this problem is much less of an issue.
As for controlling it, one project i have always wanted to try is to rig up a set of pontentiometers or bend sensors to actively measure the joint angles of my arm, then duplicate them on the prosthetic. Also, with a glove, I could possibly shake someone’s hand with a robot!
Still mulling this over, but it’s something I definitely want to try.