Re: [ompl-users] Straight line (in config space) path?

[Patrick B. <bee...@gm...>]

Thanks.  I know it is the case that some other planners (RRT for example)
will do this as well.

Basically, I think many of my issues could/should be improved if the
optimizing planners had a mode switch that would allow them to return
immediately upon finding an answer.  That is, if I've already built out a
roadmap for PRM*, there's no need to keep optimizing at some point. I want
to flip a switch that says just give me an answer.

Similarly, another "optimization" to PRMs and their derivatives in general
would be to always add a graph/roadmap edge between any start/goal pair fed
into the planner (for PRM this would require collision checking, but for
Lazy would be evaluated later)   This could/would create "super-highways"
in the node graph which might be beneficial long term, but also when paired
with the mode I mention above would basically allow many planning problems
to return back a solution immediately if the start/goal is the shortest
known solution.

I realize that the two things above don't exist, and that the second item
is more of a "Research" item that would require evaluation to prove out my
hypothesis.  But having a mode switch to turn off "exploration" for some of
the OMPL planners just seems like a practical thing to have.  Otherwise,
these planners that run for the full timeout become largely unusable in
time-sensitive contexts.

On Wed, Jun 24, 2020 at 9:14 AM Jonathan Gammell <ga...@ro...>
wrote:

> > Message: 1
> > Date: Tue, 23 Jun 2020 10:01:21 -0500
> > From: Patrick Beeson <bee...@gm...>
> > To: omp...@li...
> > Subject: [ompl-users] Straight line (in config space) path?
> > Message-ID:
> >       <CAEt3jgKmkb4JJOmbNT=
> 5jA...@ma...>
> > Content-Type: text/plain; charset="utf-8"
> >
> > In using OMPL via MoveIt!, I'm noticing that OMPL often does a whole lot
> of
> > work to plan even when there's a straight line in configuration space
> from
> > the start to the goal.  So it'll find a solution of let's say 6-10 points
> > then that gets shortened down to 2, the start and goal.
> >
> > In asking about just checking the 2 point plan of start/goal for
> collisions
> > in MoveIt! (
> >
> https://answers.ros.org/question/355617/can-moveit-first-check-shortest-path-before-planning/
> ),
> > the short answer was "your planner needs to support it".  So, now I'm
> > asking here, does OMPL support this functionality as an option prior to
> > running plans, and is this exposed via the MoveIt! API into OMPL?  If so,
> > I'm missing where this is documented.  If not, is there a viable solution
> > to do this in OMPL?
>
> Hi Patrick,
>
> I am not personally aware of such a feature in OMPL in general, but some
> specific planners do behave this way as a result of their design. I'm
> specifically thinking of planners that use heuristics to inform their
> search in order of potential solution quality, such as BIT* and AIT*.
>
> You can find them in OMPL under geometric/planners/informedtrees/
>
> Full disclosure, these are planners I've worked on. They haven't been
> tested exhaustively with MoveIt!, but I believe they do work with it.
>
> Best,
> Jon
>
> --
> Jonathan Gammell, Ph.D.
>
> Departmental Lecturer in Robotics
> Estimation, Search, and Planning (ESP) Research Group
> Oxford Robotics Institute (ORI)
> Department of Engineering Science
> University of Oxford
> office: +44 1865 613082
> skype: jdgammell
> https://robotic-esp.com/gammell
>
>
>
> _______________________________________________
> ompl-users mailing list
> omp...@li...
> https://lists.sourceforge.net/lists/listinfo/ompl-users
>