Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
SAF Archives  search ask the scientists in the classroom cool science
scientists from previous shows
cool careers in science
ask the scientists
PREVIOUS SCIENTISTS

Photo of Jerry Pratt Jerry Pratt as seen on Natural Born Robots: Body Builders

Click on Jerry's photo to read a brief bio.



q Can you please explain how the legs are powered? (Asked by several viewers)

A The legs use DC electric motors, the same ones that are used in model airplanes. In fact, the ones we use are made by a company called Astroflight and are intended for model airplanes. The motor spins too fast though, so it has to be geared down. To do that we use a ball screw which is a device that moves back and forth slowly when you spin it quickly. The motion is then forward-backward instead of round and round. We then connect the forward-backward motion to a steel cable that goes to one of the joints (hip, knee, or ankle). There is a motor for each joint (6 total). All of them are located up in the body, so the legs are really light.


q What is the hardest part of designing and building a leg that works like a human's? (Asked by several viewers)

A One of the hardest parts is to figure out what to use instead of muscles. Muscles, like most of the "machinery" that nature uses is very impressive when compared to the machinery that engineers can choose from. Your arms and legs are very compliant. If you relax, they'll flop around. If you bump into something, you bounce off it. Many old robots aren't like that. If they bump into something, like a wall, they'll go right through it. In order to make a robot that walks like an animal or human, it is important that the limbs can be floppy so if something unexpected happens - like hitting a bump, the robot doesn't fall down or break itself. In our robot, we try to make the motors behave more like muscles by adding springy-ness to them. We add springs between the motors and the joints on purpose, so that if we want, we can let the joints be compliant.

The other hard part is controlling the legs. After the robot is designed and built, it takes a lot of hard work to figure out how the legs should be controlled in order to balance and walk. For example, if you want to go faster, what should you do? People can speed up and slow down without thinking about it, but when you really do start to think about it, it is hard to explain what you are doing. Go ahead, try it for yourself! Walk slow and then fast. Think about how you push off harder with your foot at the end of the stride, how you swing your leg faster when you walk faster, how you take longer strides as you walk faster, etc. We ask ourselves these questions in order to come up with ideas for how to control the robot. Then it takes a lot of experimenting on the robot to really see which of the ideas are right. What makes it really tricky is that animals and humans use a lot of methods to balance, speed up, swing their leg, and so on. For the robot we try to discover all the tricks and use them all.



q If robots could be as fast and limber as humans, when do you foresee that happening? (Question sent by Diana, Cupertino, CA)

A I'm not sure if robots will ever be as fast and limber as humans. Consider a ballerina dancer, football player, piano player, high jumper, skydiver, skier, soccer player, or swimmer. Many people can do all of those things (but probably only a couple well). A robot may be able to do certain things better than a human, but I don't think a robot will be able to do so many things so well.

A lot of these activities require lots of learning which robots are very bad at right now. They all require good actuators (muscles) and sensors that are much better than what is available for robots. It may be possible that the best way to do these activities is the way nature does them. If a robot is to be as athletic as a human, maybe it will need to use muscles and other stuff that nature uses. The best way to make these things is to grow them and the best way to grow them is to use cells, and DNA, and all the stuff nature uses to grow them. If we end up doing that, then will the robot be a robot or will it be an animal?

For right now though, a more practical question will be when will robots be good enough that they don't fall down and burn up when they're in a home or office. That might happen within 50 years but I don't think the robots will be too smart for at least 200 years.

I think a more practical use of robots is too aid a human, rather than replace them. Imagine a human that can run as fast as a cheetah by wearing a suit that helps him run. The human would do all the tough thinking and be the intelligence. The suit would only do the things that robots are good at. Most inventions that have been useful do exactly this. For example, computers don't replace people, they only make them smarter by enhancing the abilities of the person using the computer. I think the same thing will be true with robots. We're already seeing this with robots like the under-water robots that helped divers explore the Titanic.



q Could you program the robot legs to walk different ways or at different speeds? Or to limp like it was hurt? (Question sent by Mightybug)

A Yep. The robot can move really slowly or kind of fast. The fastest that Spring Flamingo can walk is about 1.5 miles per hour which is still pretty slow compared to humans. We're working on figuring out how to make the robot walk faster. We can have the robot take really long strides or really short strides to make it look different. I think right now the robot does look like it's hurt since it's so hard to make it look good. The real question is when will we be able to make it look like it's comfortable and confident? We're working on it!




 

Scientific American Frontiers
Fall 1990 to Spring 2000
Sponsored by GTE Corporation,
now a part of Verizon Communications Inc.