Subject:
|
Re: Robotic simulators
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 26 Apr 2005 00:48:08 GMT
|
Original-From:
|
Steve Baker <SJBAKER1@AIRMAIL.NETstopspammers>
|
Viewed:
|
847 times
|
| |
|
|
dan miller wrote:
> I'd be interested to hear about your experiences. The codebase I'm looking
> at is called Gazebo (http://playerstage.sourceforge.net/gazebo/gazebo.html)
> -- it sits on top of ODE, and seems to do a pretty good job of simulating
> simple wheeled robots, at least.
The thing I hate about ODE is that it doesn't like you giving it real-world
units.
If I have a little kart model (such as the demo model that comes with ODE),
and it has a mass of about 1 unit, with appropriate spring stiffnesses, etc,
everything looks good and works well.
However, if I enter the actual mass of my car (around 1200kg say) - and the
actual stiffness of it's springs (as obtained from the manufacturer) - then
drop the car from 10 centimeters above the ground - it bounces into orbit!
When I complain to the ODE developers about this, all they say is to scale
back the units to be around unity...which is all very well - but having put
the mass at unity, you have to do some pretty significant math to get appropriate
spring stiffnesses, etc.
> I'm surprised to hear that. While obviously it wouldn't be perfect, it
> seems to me that the granular, quantized nature of Lego parts would make
> calculating the behavior of various linkages, gearing, etc. pretty
> straightforward.
Yeah - but it's hard to measure stuff like the friction of axles, the
rotational inertia of things as weird-shaped as Lego parts...etc.
> I'm not saying it wouldn't be a considerable undertaking; just that it seems
> conceptually possible given the present state of the art in simulation
> techniques.
It's *conceptually* possible - I agree. But whether it's possible in practice
is a lot more doubtful.
Complex systems are rarely simulated from the ground up with any degree of
success. I work in Flight Simulation - we never take the shape of the model
and do finite element mesh analysis to determine rotational intertias, virtual
wind tunnel experiments to calculate drag and theoretical engine thrusts calculated
from first principles. We take actual wind tunnel data from actual aircraft
actual measured aircraft performance figures - engine parameters taken from
real measurements and so forth.
That's fine when you are simulating a whole real world thing like a car or a plane.
But taking something as infinitely reconfigurable as Lego means that you cannot
usefully measure the real world system - you'd HAVE to calculate everything from
first principles - which (as I've explained) is notoriously difficult to match
to the real world.
> For the most part, Lego creations operate in a near-static
> regime; you wouldn't have to worry about higher-order dynamic effects (in
> other words, the inertia of parts is usually negligible compared to the
> constraints the parts impose on each other... things aren't bouncing around
> like in pinball machine.)
Inertias may be a small matter - but frictional forces are not.
> Connected blocks would be treated as rigid bodies. Almost all linkages are
> what are called 'hinges' in ODE -- one degree of freedom around an axis.
> Gears could be dealt with explicitly, or as wheels in contact with infinite
> friction. Once the motor kicks in, the operation of the machine should
> progress pretty much deterministically.
Yeah but things aren't like that with real Lego.
Try this (for real - with actual Lego).
Take a couple of 24t gears - set them up on axles that are spaced 2.5 studs
apart horizontally and offset 4 plate-thicknesses vertically. Compare how
easily these rotate versus the same two gears spaced 3 studs horizontally and
offset one plate vertically. Both pairs seem to 'fit' - but in reality, the
first pair are 0.032 studs too close together and the second pair are 0.027
studs too loose.
Most robot builders will use both of those gear setups as if they were correct,
but the frictional differences are very significant.
But don't stop there - read this:
http://www.sjbaker.org/steve/lego/motor_speed.html
...especially the part about 'build accuracy'.
> Anyway, seems like a cool idea. If anyone has a serious interest in this,
> pls contact me -- danbmil99 at yahoo dot com
It is a cool idea - but the realisation of it would be an enormous effort
IMHO - and the value of it as a tool for robot builders would be limited.
---------------------------- Steve Baker -------------------------
HomeEmail: <sjbaker1@airmail.net> WorkEmail: <sjbaker@link.com>
HomePage : http://www.sjbaker.org
Projects : http://plib.sf.net http://tuxaqfh.sf.net
http://tuxkart.sf.net http://prettypoly.sf.net
-----BEGIN GEEK CODE BLOCK-----
GCS d-- s:+ a+ C++++$ UL+++$ P--- L++++$ E--- W+++ N o+ K? w--- !O M-
V-- PS++ PE- Y-- PGP-- t+ 5 X R+++ tv b++ DI++ D G+ e++ h--(-) r+++ y++++
-----END GEEK CODE BLOCK-----
|
|
Message is in Reply To:
 | | Re: Robotic simulators
|
| --- Steve Baker <sjbaker1@airmail.net> wrote: [3D Lego simulator] (...) I'd be interested to hear about your experiences. The codebase I'm looking at is called Gazebo ((URL) it sits on top of ODE, and seems to do a pretty good job of simulating (...) (19 years ago, 25-Apr-05, to lugnet.robotics)
|
4 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
