Robot critters by Rebecca Roberts

Picture your favorite robot. Whether it's Robbie from "Lost in Space" or "Star Wars'" C3PO, chances are your robot looks more or less like a human. But there's a growing trend in robotics called biometrics ­ where scientists create robots that look and act like other animals. As "The World's" technology correspondent Rebecca Roberts reports, the goal is to have robots do things humans aren't very good at, like walk the ocean floor, or even fly.

Ayers: One of the things that lobsters do that most other animals don't do well is walk in any direction. They can walk forwards, backwards, laterally the two directions. They can switch between walking directions on a step-by-step basis. That gives them incredible ability to maneuver through very irregular areas: rocky bottoms, areas where there's seaweed, stuff like that.

That's something current underwater robots, like the ones used to retrieve the black boxes from the Egypt Air crash, aren't very good at. That's one of the reasons those robots are still attached to a boat with heavy, cumbersome cables. A truly autonomous underwater robot would be extremely useful, especially for the Navy, who would like a robot to retrieve unexploded mines from the ocean. But a de-mining robot would have to negotiate irregular ocean floors, and stay on course in turbulent waters. So Ayers and his team looked to nature for creatures that had already solved these problems, and came up with the lobster.

It's called "biomimetics," the imitation of a natural phenomenon in robotic form. And the scientists who practice it say it's just like using a really good library -- why start from scratch to come up with answers for problems already solved through millions of years of evolution?

John Kumph: I saw a sign in the hallway when I was a senior at MIT that said "come work in a robotic fish." I laughed and said that was a great idea .

John Kumph, now a graduate researcher at MIT, took the challenge, and is now building a robot that looks like a fish -- specifically, a pike. He's hoping robopike will teach scientists the secret of achieving both speed and agility.

Kumph: You can build a vehicle that's in the water ­ it's a long tube, it has a propeller on the back and it will go very very fast through the water. But it doesn't turn well. You can build a vehicle with thrusters all over it, it looks like a refrigerator, and it will turn very well, but it doesn't go straight very well.

Both robopike and robolobster are designed for jobs that are dangerous or difficult for humans. But across the country, at the University of California at Berkeley, scientists are building a robot designed to do what's impossible for humans: fly.

Berkeley professor Micheal Dickinson is working on robofly, designed to mimic a standard house fly. The goal is to create a flying machine totally unlike an airplane or a helicopter.

Dickinson: The main difference is that insects flap their wings back and forth. And unless you're very unlucky, airplane wings are rigid and fixed, move in one direction during a flight from Chicago to Tokyo.

Robofly would be able to stop, start, and change directions and altitudes easily, less like a plane and more like an insect. But flight isn't the only fly talent Dickinson is trying to imitate in robofly. Flies are also great information processors, taking in visual information much faster than humans. And flies, unlike robots, are incredibly resilient.

Dickinson: The number of times that you've tried to swat a fly and sometimes you clearly hit it, you know you hit it. Like Mark McGwire with a rolled up paper. And it slams against the wall, falls on the floor, and 30 seconds later it kind of shakes, slaps its head, and it's flying again! This is extraordinary! Imagine doing that by any human-made mechanical device.

Robofly, which, like robopike and robolobster, is being partially funded by the US Navy, could eventually be used for stealth surveillance, or search and rescue applications. If Dickinson and his team can mimic a fly's uncanny ability to find a warm, sweaty human to land on, robofly could find trapped earthquake survivors. The goal is to make robots that can do the tasks that are too dangerous or impossible for humans. If you find the animal that's already good at that task, you have a model for your robot.

But any practical application is still years away. Robofly has yet to fly. And although robopike can swim, it doesn't have nay sort of intelligence or autonomy. Robolobster is on schedule to be finished by February 2001. But one of the toughest hurdles facing robotic critters is one of the most mundane: unlike their counterparts in nature, these robots need batteries. And even the most high-tech batteries are still too big and bulky and run out too soon to make robotic animals anything more than an interesting experiment.