answers cntd.

> >  If you are cerful you can switch between algorithms or their
> >  parameters even during motion execution. The problem is to have the
> >  current values in both algorithms, e.g. especially the integral
> >  component, but also the previus value for the sake of
> >  differentiation.
> 
> Have you ever seen an implementation of this on-line switching? Do you
> think there is a need for it?

We used the switch of algorithm for rapid testing of several algorithms, but I doubt if you will really need that otherwise. If you would need this in a mature device you would rather implement a complex algorithm consisting of several simpler ones. The current solution is good for research purposes only - at least this is how I see it currently.

> 
> I can think of one (or rather, a whole family of cases), namely hybrid
> control: one could leave the responsibility of the switch robustness
> to the programmer, or one could implement sub-optimal but robust
> default switching behaviours. But I don't have a clear view on the
> matter, let alone any experience. But I do want to leave itopen as an
> option.

Seems reasonable.

> 
> > > - how do you extend the Move and Wait commands for sensor-driven
> > >   commands, such as force control or visual tracking?
> >
> > You don't have, as sensor interaction is included in the Move or
> > Wait instructions. Please note that you deliver both the list of
> > sensor images used and the generator/condition as arguments. The
> > list determines to which sensors you want to talk and the generator
> > determines how you use the sensor readings in motion
> > generation/modification.
> 
> Okay, I see. I was a bit misled by the name `generator', because that
> doesn't suggest (to me at least) its functionality as a servo control
> loop (which it is if you take sensor information into account, I
> guess).

Actually this is a an abbriviation of motion generator. This kind of generator can switch servo algorithms, but otherwise is not responsible for servo control. The sensor input implements high level feedback - not the servo level one. Usually those generators generate trajectories, which can be modified by sensor readings. To be strict they generate the next target position. As our servo sampling rate is either 1 or 2 ms the generator usually performs at the rate of 6 or 8 ms (for the 2 ms sampling rate), so its influence on the trajectory is fairly rapid.

Cezary
> _______________________________________________
> Orocos mailing list
> [email protected]
> http://mail.mech.kuleuven.ac.be/mailman/listinfo/orocos