Monday, July 24, 2006

Biped Seeks Cables

Here’s the first robot I tried to built. This biped has four degrees of freedom per leg. It has an ankle, knee, and two degrees of freedom in the hip. I took a design where the actual load on the motors while standing up would be as low as possible. Here’s a picture of the biped sitting on a chair.

first biped bNXT

As you can see, the robot is quite large. My main problem with this robot is that I don’t have cables long enough to connect the tip of the toe with the controller at the hip. That’s really too bad. I’ve been hitting the “no cables” nail quite a few times before, but this is really a nice demonstration why a crimp tool makes more sense that predefined lengths of cables.

In later robots, I found that I would have preferred to use more ultra-short cables.

Filip

10 Comments:

At July 25, 2006 3:54 PM , Anonymous mr. M. said...

Wow, this thing really walks? Looks very unstable :)

There shouldn't be problem to shorten your cables, i think... ?

 
At July 25, 2006 5:15 PM , Anonymous david levy said...

Can't you just purchase longer cables? Aren't these just standard rj-11 jacks ?

 
At July 25, 2006 11:05 PM , Anonymous msim said...

(if possible) what if you could put two of the NXT bricks near the knees, or could be the knees.

 
At July 26, 2006 8:09 AM , Anonymous Anonymous said...

Long cables aren't a problem. Just use RJ12 cables, snap the spring clip off and then put 2-3 layers of masking tape on 3 sides of the connecter (not the bottom). You will then find that the plugs jam into the NXT sockets.

NOTE: Just make sure you make your RJ12 cables the correct way around.

 
At July 26, 2006 10:16 AM , Anonymous Filip said...

Yes, the robot is very unstable, especially with Lego servo's. They are under microprocessor control, and the delay in response time causes issues. But the main thing holding this robot back is the lack of a tilt sensor (accelerometer), for dynamic adjustment.

To shorten my cables, I'd also need a crimp tool for the Lego set. The Lego connectors are unique. They are not rj-11's. Thanks for the tip on converting them. Philo also has published a conversion method. But I'm still hoping to see an announcement by Lego in the near future.

Putting the NXT sets lower reduces the freedom I have with the legs. Still, putting them at the knees might work, I didn't fully investigate it. Right now, the project is on hold until I have at least a two-directional accelerometer, that can detect whether it is falling over.

 
At July 26, 2006 7:30 PM , Anonymous Anonymous said...

I can't really see how that thing could get off the ground. Can a Lego motor (even with the plastic gears) lift 3 NXT bricks?

Anyway, nice creation! I instantly imagined a NXT fisherman sitting on a dock by the bay...

Here's that accelleraion sensor you require...

Where I heard about it:
http://nxtasy.org/...

Direct Link: (bottom of page)
http://www.mindsensors.com/...

 
At July 29, 2006 9:16 AM , Anonymous Anonymous said...

Your biped will never walk, at the actual state of things.
Servo motors need to be reduced with geartrains, and there's no hope to control that number of servos with a mere accelerometer.
The weight of NXTs must be distributed better, even if main problems are closed loop control and structure weakness.
I'm waiting to see how can such a complex system be managed with G-language!

 
At July 31, 2006 6:49 PM , Anonymous Filip said...

Anonymous, these are valid remarks.

There is a tradeoff between motor power, speed and battery usage. Lego has proposed a tradeoff in the current NXT motors, and these tradeoffs make building this biped hard. There are clear limits to the power these NXT motors can provide. Gearing the motors down is probably not the best solution, since that reduces the speed. Without sufficient speed, the robot can never stay on its feet.

An alternative solution is to put two motors side by side, working together to provide the requested power. Assuming they are properly calibrated, this can be done.

I don't see why I would limit myself to a single accelerometer, or to the NXT-G. If NXT-G proves a problem, I can always use NBC or RobotC.

Even so, the robot may still never walk. This robot is nothing more than an experiment. But it would sure be fun to see what it CAN do, and instructive to learn what the main remaining issues are.

Filip

 
At August 25, 2006 3:53 PM , Anonymous Drab said...

You can find a tilt sensor (leveler) at http://www.vernier.com/nxt/. Their leveler is actually a vernier Low-g Accelerometer that can be purchased on their site.

Secondly, if the motors can handle the weight and you find the extended cables, why not move two bricks to the feet for increased stability. And, if it's not too suggestive, leave the third where it is.

 
At August 25, 2006 4:11 PM , Anonymous Filip said...

Great link, thanks.
Filip.

 

Post a Comment

Subscribe to Post Comments [Atom]

Links to this post:

Create a Link

<< Home