[ODE] Picking objects

Thu Jul 15 11:18:15 MST 2004

>  when you say follow the 'Proxy' at best => it means to 'Constrain' the physical object to the position of the proxy as in a joint (LCP).

Right yes or even just a plain vanilla set of equality constraints - LCP being mainly for the unilateral stuff eg contact and so forth.

But my point was not supposed to be about how to implement constraints - more about how to drive a robotic thing interactively - something that needs to be done quite carefully because simulated robots tend to explode and the real thing will either bash into things or just cut power (you hope) when your software starts making unreasonable demands.

At least one extra "object" or layer, if you like, between you (ie user input) and your mechanism, is a way to simplify matters. The state of the proxy at any time reflects user intention, the difference in state of proxy and state of mechanism is your "error" and the controller (dynamic, kinematic, whatever..) attempts to minimise that "error" - by moving.

>  When you say 'Clamp' acceleration and velociy it means you are playing with State params as per your 'taste'

Well, yes as per some set of (more or less arbitrary) criteria - call it a sanity check/override. If you understand your model it's often possible to derive operational limits and ranges. A silly example, a projectile falling under gravity against some form of air-resistence or whatever will have a "terminal velocity", any output far above that would signal some form of breakdown in the physics pipeline. A policy of clamping in this instance may well mean the difference between complete melt-down and a slight "glitch in the matrix".

> => in other words fuzzyfication.

possibly.

:)

> -----Original Message-----
> From: Manohar B.S [mailto:sciphilog at yahoo.com]
> Sent: 14 July 2004 14:03
> To: Bob Dowland; ode at q12.org
> Subject: RE: [ODE] Picking objects
> 
> 
> 
> Ya, spring-damper works nice when tuned properly. But
> I didn't get the cear idea what you have done with the
> Robot Arm. Can you explain with a small code.
> 
>  when you say follow the 'Proxy' at best => it means
> to 'Constrain' the physical object to the position of
> the proxy as in a joint (LCP). Some people do this by
> connecting the 'proxy' and the ODE body with a
> spherical joint (or any other joint). This should work
> fine I guess.
> 
>  When you say 'Clamp' acceleration and velociy it
> means you are playing with State params as per your
> 'taste' => in other words fuzzyfication.
> 
>  -- Manohar BS.
> www.scientifik.org
> 
> --- Bob Dowland <Bob.Dowland at blue52.co.uk> wrote:
> > Guys,
> > 
> > some time ago I had to simulate a robotic arm and
> > control the pos/ori of the end effector using output
> > from a 6d pos/ori tracker - lots of fun. What I did
> > with that might be of interest and might not but
> > there we are.
> > 
> > I used what you might call a "reference proxy" which
> > responded perfectly (ie kinematically not
> > dynamically) to the user input. The motion of this
> > proxy was totally free and unsullied by the ravages
> > of physics/collision response - it could go through
> > walls, outreach the arm and so on.. The arm then
> > just "did the best it could" to follow it. Where
> > "doing the best it could" meant not violating it's
> > various dynamic and geometric constraints. This
> > worked very nicely.
> > 
> > As far as spring and damper mechanisms go the main
> > problem seems to be the tuning issue. In particular
> > in graphical sims the spring (and underlying
> > metaphor :) ) will often get "stretched" far beyond
> > any reasonable physical limits / dynamic operating
> > range - so in a sense the model breaks down. One
> > awful hack that works quite well is to treat the
> > spring damper as a black box, tune stiffness and so
> > forth to taste (so that things are good where you
> > mostly need them) then maintaining a very clear idea
> > of what velocity/acceleration ranges you want to
> > allow just clamp the outputs.
> > 
> > Bob.
> > 
> > > -----Original Message-----
> > > From: Manohar B.S [mailto:sciphilog at yahoo.com]
> > > Sent: 30 June 2004 10:55
> > > To: Olivier Parisy
> > > Cc: ode at q12.org
> > > Subject: Re: [ODE] Picking objects
> > > 
> > > 
> > > Hello James,
> > > Hello Olivier,
> > > 
> > > > It's sad we have to get to such tricks as soon
> > as
> > > > model control is needed... Perhaps someone who
> > > > developped a stable servo example could
> > contribute
> > > > it in ODE distribution someday ? I know it would
> > > > solve most of my animation/interaction needs !
> > > 
> > > Well, I think there's a serious need for model
> > control
> > > or Motion controller interface with the physics
> > > engines (which ever it be). Well, oscillations are
> > > part of the physics and the physics engine is
> > happy
> > > becos it is honestly doing what it is supposed to
> > do.
> > > We always expect some thing that we get convinced
> > of
> > > due to our past experience with the nature
> > > (irrespective of right mathematical
> > justification).
> > > 
> > >  At this time I am not very much informative about
> > how
> > > to make a very "convincing" and "Accurate" motion
> > > control interface but I believe that Fully logic
> > based
> > > contoller is the best option. But the requirements
> > are
> > > like this:
> > > 
> > >  1. The Physics body/geom should follow the
> > position
> > > and orientation of the Motion sensor as quickly as
> > > possible (with accuracy).
> > >  2. Oscillations should be minimum or atleast
> > > invisible.
> > >  3. Forces applied on the physics body/geom should
> > be
> > > clean enough to be applied to the forcefeedback
> > > device.
> > >  3. The algorithm should be computationally lite.
> > > 
> > > ___________
> > > Fuzzy Spring-Damper will have variable spring
> > konst
> > > and variable damping konst, depending on the rule
> > > table entries.
> > > 
> > >  I would be happy to dedicate time for research in
> > > this like.
> > > 
> > >  -- \
> > > Manohar BS
> > > 
> > > 
> > > 
> > > 		
> > > __________________________________
> > > Do you Yahoo!?
> > > Yahoo! Mail is new and improved - Check it out!
> > > http://promotions.yahoo.com/new_mail
> > > _______________________________________________
> > > ODE mailing list
> > > ODE at q12.org
> > > http://q12.org/mailman/listinfo/ode
> > > 
> > 
> > _______________________________________________
> > ODE mailing list
> > ODE at q12.org
> > http://q12.org/mailman/listinfo/ode
> > 
> 
> 
> 
> 	
> 		
> __________________________________
> Do you Yahoo!?
> New and Improved Yahoo! Mail - 100MB free storage!
> http://promotions.yahoo.com/new_mail 
> 