[chbot] Balancing robots

[Charles Manning]

On Tuesday 21 October 2008 11:27:27 Michael Pearce wrote:
> > This is a level of stability and control not achievable with simplistic
> > PID or Fuzzy controllers.
>
> Actually the Segway used a good number of dsPICs to do it all.... so
> not as much processing power as you may have presumed.... (Note: dsPIC
> is _NOT_ a dsp- it is a 16 bit digital signal Controller up to 40MIPS)

I thought the Segway actually used a TI DSP chip, but perhaps that was an 
earlier model. See a press release at: 
http://www.embeddedstar.com/press/content/2003/4/embedded8283.html

You can implement useful algorithms on an 8-bit or 16bitter or anything else, 
though obviously more expensive algorithms need more grunt.

The nBot achieves good stability with a 68HC11  (1980s 8-bitter), this 
includes using a complementary filter as well as other controls.

The basic nBot algorithm (first generation) is at: 
http://www.geology.smu.edu/~dpa-www/robo/nbot/bal2.txt but the final version 
(4th gen) uses far more sophisticated input filtering (the complementary 
filter).

http://www.youtube.com/watch?v=uaaNUMwI3xg

The real test of the control algorithm, IMHO, is how it handles transitions. 
In the video you can see how well it handles going over the crack in the 
concrete. I have seen other videos of the nBot going up a rough four wheel 
drive forestry track.

The nBot balancing is very good when you consider how compact it is. A large 
item like a Segway is far simpler to balance. To convince yourself of this, 
try balance a pencil on your finger. Then try balance a broom stick. 

Fuzzy controllers can be set up to work in many different ways. Depending on 
configuration they can mimic PIDs and state space controllers, as well as 
other controllers.

Ultimately any control technology is only as good as its tuning and 
environment. I have seen a state space controller being handsomely beaten by 
a simple PI (no D) controller because the state space controller was too 
fragile and hard to tune. The PI controller was far more robust and easier to 
tune.


>
> Was definitely in the 2007 version.... may be different now.
>
> Mike
>
> On Mon, Oct 20, 2008 at 1:11 AM, John Stowers <john.stowers at gmail.com> 
wrote:
> > This was created by Dean Kamen, i.e. Mr Segway. I expect it uses some of
> > the control technology inherited from that project.
> >
> > Presumably, like the segway, it uses one or multiple voting, optimal
> > state space controller/s.
> >
> > This is a level of stability and control not achievable with simplistic
> > PID or Fuzzy controllers.
> >
> > John
> >
> > On Mon, Oct 20, 2008 at 1:34 AM, Charles Manning <cdhmanning at gmail.com>
> >
> > wrote:
> >> Pretty neat.
> >>
> >> I think there are a few things which really help keep this thing stable:
> >>
> >> 1) Moment of inertia.  A heavy item placed relatively far from the
> >> wheels is far easier to balance than a light one. That's part of the
> >> reason Hanno has the batteries on the top of the pole of his robot .
> >> Having an object the size of a human helps.
> >> 2) Relatively unlimited budget. Hanno is trying to make a cheap consumer
> >> item while these flash wheelchairs cost many thousands.
> >>
> >>
> >> Some horizontal reference, typically accelerometers because they are
> >> small, cheap and reliable, need to be part of any gyro-based solution
> >> because gyro sensors give you the rate of rotation and you need a
> >> horizontal reference to cancel out any offset (normally done in a Kalman
> >> filter). If you don't cancel out this offset then the machine would seem
> >> to be slowly rolling.
> >>
> >> -- Charles
> >>
> >> On Sun, Oct 19, 2008 at 5:23 PM, QtuTrains <Howard at qtutrains.com> wrote:
> >>> If you think Hanno's balancing robot is interesting just take a look
> >>> at this:
> >>>
> >>> http://www.powerchair-review.fsnet.co.uk/ibot-4000-transporter-full-
> >>> review.htm
> >>>
> >>> The manufacture's site is http://www.ibotnow.com/ but I couldn't get
> >>> it to respond when I looked.
> >>>
> >>> Its a wheelchair that can rise onto two wheels and balance with a
> >>> person sat in it.
> >>>
> >>> It can also climb stairs!
> >>>
> >>> We saw one today at the Canterbury museum (not an exhibit!) and the
> >>> owner said it was the only one in New Zealand.
> >>>
> >>> It looked amazingly stable - much less oscillation to keep upright
> >>> than Hanno has achieved thus far. Naturally it uses gyroscopes for
> >>> balance, perhaps accelerometers too but the owner wasn't too much
> >>> into the technical details.
> >>>
> >>> The point of balancing is improved manoeuvrability and visibility
> >>> through increased height which was obviously useful in the museum.
> >>>
> >>> Howard
> >>>
> >>> _______________________________________________
> >>> Chchrobotics mailing list
> >>> Chchrobotics at lists.linuxnut.co.nz
> >>> http://lists.ourshack.com/mailman/listinfo/chchrobotics
> >>
> >> _______________________________________________
> >> Chchrobotics mailing list
> >> Chchrobotics at lists.linuxnut.co.nz
> >> http://lists.ourshack.com/mailman/listinfo/chchrobotics
> >
> > _______________________________________________
> > Chchrobotics mailing list
> > Chchrobotics at lists.linuxnut.co.nz
> > http://lists.ourshack.com/mailman/listinfo/chchrobotics