THE VIBRATING eyes of jumping spiders have inspired a new breed of vision sensors that could give the next generation of Mars rovers sharper eyesight, say researchers in California. As a result, the roving robots will need less computing power, so they'll be much lighter and will use less electricity.
Today's robot-vision systems sense images focused onto an array of picture elements on a microchip. Each picture element, or pixel, is generated by a photoreceptor that converts light intensity into an electrical signal. The sensors generate huge quantities of data many times per second - so they need complex computer circuitry to process the information. But this adds weight.
To cut down on the processing power required, Oliver Landolt, Ania Mitros and their colleagues at the California Institute of Technology in Pasadena are turning to nature for help. Biologists have long known that the tropical jumping spider, which vibrates its thin, long retina back and forth to see more clearly than would otherwise be possible.
Landolt and his colleagues wondered whether vibrating an electronic image sensor could also create sharper images. So they built an array comprising 32 rows of 32 pixels each on a silicon image-sensor chip and clamped a metal frame onto it. Next, they mounted a lens inside the frame, attached by four springs. The lens focuses an image onto the chip. When the unit is shaken, the lens starts vibrating at the springs' resonant frequency - about 300 hertz - moving the image over the pixels.
The distance from the lens to the surface of the chip never changes, so the image is always in focus. The net effect is as if each pixel is moving elliptically, sampling a larger area of the image than if it were stationary. Each pixel generates a train of digital pulses from which a signal processor extracts features of the image - using knowledge of the position of the springs.
|To see why moving pixels are better than stationary ones, consider the line that signals a transition from light to darkness (see Diagram, bottom right). Immobile pixels can only tell that the transition is somewhere in between them. But if the pixels are moving, they cover the region between them, and will detect precisely where the change in light intensity occurs.|
Although smaller than a matchbox, the vibrating sensor's images are almost as sharp as those of an immobile sensor with 256 by 256 pixels, says Landolt. "This is a good start to doing interesting things with a robot, such as finding where a door is, or identifying the orientation of edges between light and darkness."
Christopher Assad, who works on biologically inspired planetary robots at NASA's Jet Propulsion Laboratory in Pasadena, agrees. "This would be great," he says. "The really elegant part that I like is that if you mount this on a small rover, the rover's going to have vibrations just from its movements and you can use that wasted energy to power the sensor's movements. It's a very elegant way to win big on power and allow you to do smart imaging."
Landolt's team, whose research was funded by DARPA, the Pentagon's research outfit, revealed its vibrating camera at last week's Advanced Research in Very Large Scale Integration conference in Salt Lake City, Utah.
Anil Ananthaswamy, San Francisco
31 March 2001
© Copyright New Scientist, RBI Limited 2001