On Wed, Jun 5, 2013 at 4:49 PM, Adolfo Rodríguez Tsouroukdissian
<
adolfo.rodriguez@pal-robotics.com> wrote:
> It makes a lot of sense in my mind to have a more generic interpolator
> message and code that leverages it. As an obvious non-joint usecase, we have
> some RGB led arrays, and it would be just great to reuse a generic nD
> trajectory interpolator to control them.
>
> @Mrinal, does your implementation provide a mechanism for specifying custom
> interpolation functions for certain fields?; for example one might want to
> interpolate orientation data represented by quaternions, whose components
> are best not interpolated independently, but using slerp.
For now, this is not specified in the message itself, since the
message knows nothing about the names of the dimensions/variables. The
low level interpolator knows about all the variable names that it
could possibly accept, since it has to distribute them to the
appropriate controllers anyway. So we deal with quaternion
interpolation as a special case at that level.
- Mrinal
(text/plain)