Shiny Doom Forum

rklee · Posted: Mon Sep 29, 2003 6:31 pm Post subject: Balancing Robots

Guest lecture today by another guy who worked on the bouncy walking robots (spring flamingo....etc). Saw a very simple no sensor no computer balancing system.
Imagine this: tiny 3"X3" base each EDGE has 1 wheel perpendicular to it (and under it) spinning to try to move the base in the direction of the edge (so with 4 of them it attempts to rip the robot apart in all 4 directions). They are all run at the same time.
Center of the base is a big hole where there's a tube attached to it very long approx 4 feet tall. It pivots in both axis at the base but extends 1" deep into the base.
When the rbot is balanced the wheels all work against each other equally and therefore it's netmovement should be zero. However in realitiy it wilkl jitter a bit.
Now when the tall pole tends to tilt towards the left for example... it pivots around the base and the lower 1" will go towards the right wheel and therefore jamming that wheel. This will cause the robot to move to the left sinc eht right motor is jammed. Same thing would hapen in the other axis. This has been demonstrated to work remarkibly well on smooth surfaces. You can actually hang weights all over the place and it will balance it very well (although the harder to balance the more space it will need).

I think Segway's cool but I think a more natural way of accomplishing tasks is better. By natural I mean it doesn't rely on a lot of sensors and it has little limitiations, it performs its job due to design rather than due to abundance of sensors. This is one reason I very much like the design of the springy robots... a lot of numerical analysis in those designs I think as opposed to an equivalent Asimo type robot. (of course Asimo's the most impressive one but I'm speaking of that type of design.... lots of sensors and control is required.... but u push it and it'll topple over... unlike the springy robots which in my opinion have a better balance system.)

Trool if you still have some of those links for the springy robots please post them. I seem to have lost them. I can also see if sometime I Can get a copy of the tapes done at CMU and MIT from the professors or find outo where I can get a copy.

Any thoughts/comments?

rklee · Posted: Mon Sep 29, 2003 6:36 pm Post subject:

Another interesting design I saw today. A rolling robot using a bike tire with two plexiglass domes on each side. It ends up being only like 5" thick or so. This was developed at CMU and uses a gyroscopic balance system. It has a gyroscope inside that spins and maintains it's orientation....
It works off of this idea. Hold a bike tire... spin it quickly and if you tilt it in the axis comming out of your chest then it will modify it's orientation with respect to the axis perpendicular to it (vertically).
In the case of this robot... when it tends to tilt say to the right... the resulting force is to rotate the robot so it turns to the right. Therefore it will balance itself.
The one here at CMU has a small motor in there with radio control so that you can actually realign the gyro (it will become misalligned eventually... obviously). So this thing is just a wheel with small curvature domes on each sode which houses the innards... and it rolls around balanced Smile

. When it falls down u can use remote control to right itself... Pretty cool eh?

I'll try to find some more info on this if people are interested also...

thoughts/comments?

Trool · Good Karma Joined: 12 Sep 2003 Posts: 1486

Nice these both sound really really cool. I agree with you too. I like the idia of a ballancing mechanism being mostly mechanical, with sensors for very special reasons. I think that makes it a lot more natural and a much easier system to deal with. The only problem i can see with mechanical designs though is that they have to be moving to keep their balance it seems. Or if not then when they do start moving it's hard for them to come back to euqalibrium and stay balanced. This is where a nice merger of the two ways may be. Sensors and electronics to start the bot and stop it, and maybe control the direction, but the movement itself be more of a mechanical springy thing. I really like both of those bots for their simplicity. I think i might still have some of those links, i'll try to find them and post on here. The videos are pretty cool.
_________________
The most exciting phrase to hear in science, the one that heralds new discoveries, isn't 'Eureka!' but rather 'hmm....that's funny.'

--Isaac Asimov

rklee · Posted: Tue Sep 30, 2003 10:29 am Post subject:

I agree with you on the merging of the two definately however I believe that it's not 100% necessary... as seen in spring flamingo.... you can have a really stationary robot... without resorting to lots of sensors... just very precise numerical analysis. Although spring flamingo was an incredibly advanced and mature robot... many years of development.

Trool · Good Karma Joined: 12 Sep 2003 Posts: 1486

I don't remmember that thing being very stable while it wasn't mooving, but i might be wrong. I though a lot of the tech in the flamingo had something to with the motion of it while it was moving....i might be getting it confused with something else though, so i'm not totaly sure...
_________________
The most exciting phrase to hear in science, the one that heralds new discoveries, isn't 'Eureka!' but rather 'hmm....that's funny.'

--Isaac Asimov

Trool · Good Karma Joined: 12 Sep 2003 Posts: 1486

Here are the link I promissed to post

--Spring Flamingo
http://www.ai.mit.edu/projects/leglab/robots/Spring_Flamingo/Spring_Flamingo.html

--Spring Flamingo and other bots walking and none at MIT
http://www.ai.mit.edu/projects/leglab/mpeg_vcd/

--Site of links to a number of bipedal bots and their websites
http://www-lag.ensieg.inpg.fr/PRC-Bipedes/English/catalogue_en.php
_________________
The most exciting phrase to hear in science, the one that heralds new discoveries, isn't 'Eureka!' but rather 'hmm....that's funny.'

--Isaac Asimov