Re: dynamic wiring

On Fri, 13 Dec 2002 [email protected] wrote:

> [email protected] wrote:
>
> > The structure of Motion control core is predeterminated, components are linked togheter by a specific mediator.
> > Why introducing a dynamic wiring system?
> > Can you explain me with an example?
> >
>
> The SmartSoft framework explicitly addresses dynamic wiring of
> components, but is meant to be applied at a much higher abstraction
> level which should not be confused with the Leuven framework.

No, but it is exactly the same thing at the conceptual level. Exactly
the same. At least, that's what I think. And so, I would welcome very
much efforts to also make the implementations and/or configuration
software as reusable as possible.


> Think of components like a mapping component, a path
> planning component and a motion control component on a mobile platform.
These correspond to our Observer, Generator and Control components at
the real-time level.

> Depending on the task to be executed one needs several data flow
> configurations between components. A first configuration might be
> mapper - planner - motion control to drive into a specific room.
> Another configuration might be vision (tracking) - motion control to
> approach or follow an object. Dynamic wiring now is the process of
> configuring these data flow dependencies at run-time to reuse
> components in different combinations for different behaviors. It is
> most needed in combination with a task execution mechanism.

That's exactly what we are targeting at in the real-time level. And
that's also the reason why I call this approach the "Software Pattern"
of (robot) control: everybody does it that way :-) The differences
being:
(i) the exact functionality in each component,
(ii) the wiring of components,
(iii) the timing constraints.
(Maybe also some other things I forget...)

Herman

-- 
  K.U.Leuven, Mechanical Engineering, Robotics Research Group
<http://people.mech.kuleuven.ac.be/~bruyninc> Tel: +32 16 322480