[ODE] RC-servo modeling

[Sergey Kurdakov]

Hello,

> The question is: do we need to model the bang-bang controller inside
> each servo, which operates on a fine time scale (microseconds?) and
> would be expensive to compute, or can we get by with using the
> built-ins of ODE, while still getting a realistic simulation of an
> RC-servo?

I think that you anyway willneed a build in controller as of timestep from my experience to model control systems it could be
relatvly low.

> We need a realistic model, because we are training articulated robots
> how to walk and we start by training them in simulation, only later
> transferring the learned walking gait to the real robot.  Without a
> realistic simulation, the transfer to the real robot may fail..

you criteria of 'realistic' is somewhat fuzzy :)

but as you will need to transfer simulations  to the real robot - you will need to simulate it's control systems as well as physics
properties. I think ode will not do tricks for you here. and it lacks 'build in' controller 'links'

BTw take a look at library http://www.cours.polymtl.ca/roboop/ maybe it will provide some hintsx

if I have some ideas - I will post them here later.

Regards
Sergey Kurdakov
----- Original Message ----- 
From: "Doron Tal" <dtal at email.arc.nasa.gov>
To: <ODE at q12.org>
Sent: Saturday, May 15, 2004 7:39 PM
Subject: [ODE] RC-servo modeling


> Howdy,
>
> We need to model an RC servo motor as part of a larger model of an
> articulated robot.  We need the RC servo model to be a good
> approximation to reality.
>
> The current RC digital servos have a bang-bang controller inside the
> servo that keeps the servo arm in one place once it is commanded to go
> there.  The bang-bang controller simply checks if the motor overshoots
> its position and if so sends the motor in the reverse direction as
> fast as possible.
>
> The question is: do we need to model the bang-bang controller inside
> each servo, which operates on a fine time scale (microseconds?) and
> would be expensive to compute, or can we get by with using the
> built-ins of ODE, while still getting a realistic simulation of an
> RC-servo?
>
> We need a realistic model, because we are training articulated robots
> how to walk and we start by training them in simulation, only later
> transferring the learned walking gait to the real robot.  Without a
> realistic simulation, the transfer to the real robot may fail..
>
> Any help would be greatly appreciated and if there is enough
> advice in the replies to aggregate, I will post a summary of the
> replies.
>
> Thanks!
> -Doron
>
> -- 
> Doron Tal
> USRA/RIACS Research Scientist
> NASA Ames Research Center
> Moffett Field, CA 94035-1000, USA
>
>
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode