Even in the world of robotic limbs, sometimes simpler is better. "Ten years ago, we kept building these anthropomorphic hands again and again, and yet we never made them function correctly," Robert Howe explains. The professor of engineering and applied sciences is Harvard's resident robotics hand expert, having pored over the subject of electronic limbs since the days of his own PhD program. A decade or so ago, however, Howe and his team decided to approach the matter differently. "We came up the idea of seeing where we could get with only one motor and very little sensing. Much to our surprise, it turns out if you get the mechanics right, one motor does pretty well, even in unstructured environments, where we aren't sure where the object is located because the computer vision isn't precise enough."

Sensing, naturally, still plays a large part in the success of a robot hand -- after all, it's tough to grab what you can't feel. Once again, Howe and his team took a simple approach to a complex problem. The group looked to smartphones for an answer, the explosion of such devices having driven down the size and price of complex electronic components. "The new generation of barometer chips that are designed for smartphones include a pressure sensor, a temperature sensor, an analog to digital convertor, high quality instrumentation amplifiers, a microcontroller and a bus interface -- and they only cost about a dollar, says Howe. "We had to convert them to touch sensors from air pressure sensors, so we tried putting rubber over the top of them. That worked, but it gave us very insensitive devices. The trick was to put liquid rubber over the chips and then pump it down in a vacuum chamber. As a result, we get very good sensitivity."


0 Comments

Peripheral Vision 013: Robert Howe on why simplicity is the key to building a better robot hand