Subject:
|
Re: balancing bot
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Thu, 30 Dec 2004 21:51:19 GMT
|
Original-From:
|
dan miller <danbmil99@yahoo.com{nomorespam}>
|
Viewed:
|
987 times
|
| |
|
|
thanks much for the links! Using a mod of Frederic Siva's code, I've managed to get my bot
somewhat stable. It wobbles around like a drunken sailor but basically stays up. I've got a
movie file of it careening around the floor for about a minute (will post on my site if anyone
cares).
What frustrates me about this project is the incredible degree of dependence between the software
parameters and the hardware. The slightest change, like batteries dying down, or my flashlight
moving a millimeter, requires a recalibration of the software to regain stability. ISTM there
should be some sort of self-calibrating routine, like a learning algorithm or something, that
would make things much more robust. I assume that's what 'real' robot people have done, or how
would their stuff ever work reliably? Obviously the Segway people got something right. I've read
they use incredibly expensive, precision parts.
I think someone mentioned doing simulations on one of the posts. I'm very interested in setting
up a simulation environment that would allow me to fiddle with some ideas on self-optimizing
systems. I've looked at Webots, which is basically a front-end to ODE -- it looks pretty scary.
I'm thinking of something more along the lines of "minimal simulation" (there are some papers on
the subject) -- where you try to capture only the most essential elements, and basically randomize
everything else (to make sure your program is robust in the real-world, noisy environment).
Enough for now, I've been cheating sleep for a week trying to get this damn thing working.
best - dbm
--- John Hay <jhay@icomtek.csir.co.za> wrote:
> Hi Dan,
>
> >
> > lately I've been playing with legway-inspired ideas. After much frustration, I'm finally • getting
> > some results using a standard lego light sensor with some custom enhancements (not taking it
> > apart, just adding a lens and a light source). At this point I think the problems are more
> > software control than hardware issues. I'm curious if anyone has been working on something
> > similar, and/or would anyone be interested in trading ideas, designs, and software? The • eventual
> > goal is to design a two-wheeled balancing robot that can be built with standard Lego parts. • I'm
> > convinced it's possible.
>
> I have also been playing with it the last week or so and got it to work.
> At the end I used Steve's legway.c program with everything except the
> balancing removed and used the raw sensor values, tweaked a bit. That
> together with two "hardware mods". I had to mount the sensor at an angle,
> and not directly downwards and do it in a darkish room.
>
> You can see photos and the modified program at:
> ftp://ftp2.za.freebsd.org/pub/lego/balance/
>
> John
> --
> John Hay -- John.Hay@icomtek.csir.co.za / jhay@FreeBSD.org
>
|
|
Message is in Reply To:
 | | Re: balancing bot
|
| Hi Dan, (...) I have also been playing with it the last week or so and got it to work. At the end I used Steve's legway.c program with everything except the balancing removed and used the raw sensor values, tweaked a bit. That together with two (...) (20 years ago, 30-Dec-04, to lugnet.robotics)
|
8 Messages in This Thread:
  
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
