There are times when I have a little bit of deja vu when hearing about the latest components, whose virtues are basically “faster, better, cheaper” (pick one or more) than their predecessors. Yes, the improvements are critical and enable new systems and circuit designs, but something's missing.
That's why I was excited to read about a new approach to an old problem, made possible by clever combination of diverse sensors. If you have ever done some star-gazing, you know that identifying “what star is that?” or “which is the xyz constellation?” is a challenge. You can use a star chart, of course, but it's awkward and frustrating. Or, you can spend about $50 and get a low-power viewer which has star charts super-imposed on the image. If you can match up the chart in the viewer with the image as you look through the unit, you can identify the stars. But in practice, using the star-finder is not easy.
Then I read about the SkyScout from Celestron, a well-established maker of telescopes and related products. It uses a different approach. You aim this $399 device (I never said it was cheap!) at the star(s) of interest, and it literally tells you (through an earphone) what you are pointing to. There's also a PC link, of course.
And how does it do this? It uses GPS to figure out where you are, a magnetometer to determine in which direction you are pointing, and an accelerometer to measure the inclination of the unit. By combining the data from these sensors, it knows where you are and exactly where you are pointing; combined with its internal sky chart knowledge, it tells you what's up there. Complicated? Yes. Clever? Very. Effective? Absolutely!
The availability of low-cost, effective sensors is forcing new ways of thinking about old problems in other areas as well. The current issue of Scientific American (October 2006) has a feature story on “Ballbots.” These are robots which do not scoot around on a pair of wheel, or tracks, or any other conventional method.
Instead, the cylindrical ballbot stands tall on a single rolling ball, about the size of a bowling ball, which is driven in x- or y-direction by associated motors and contact wheels. Think of it as a rolling-ball mouse in reverse: the mouse has two rotating sensors which detect motion of the ball; imaging that those sensors were motors instead and could drive the ball. The ballbot stand like a person, and can move through narrow corridor and doorway.
So how does it work? Without active closed-loop control, the ballbot would tip over immediately. But the robot uses internal gyros and accelerometers to sense its vertical, and correct accordingly. And when it needs to move, it deliberately leans into the direction of travel for better starting stability, then straightens up once it reaches speed.
This radical rethinking of what a robot should look like and its form factor opens up new opportunities in design and function. And it is largely due to sophisticated sensors, coupled with significant processing power, of course.