[ompl-users] Strange results from RigidBodyPlanningWithODESolverAndControls

[Andrey N. <and...@go...>]

Hi OMPL Users,

I just start learning OMPL and got couple of questions. Would really
appreciate if someone could help me. I've compiled the recent version
from Git on x86 XUbuntu. The RigidBodyPlanningWithODESolverAndControls
demo application is closest to what I need so I start playing with it.

Running the demo without any changes generates the path shown on this
image: https://goo.gl/eHW4pk . It reaches the goal, but it does not
look like a "good" solution. At least I would not drive this way
myself :-) . In particular, the movements around x=0.1, y=0.05 and
similar back and forth on the other places looks kind of strange.

I assume there are many parameters the one can tweak to achieve better
result, but I did not found any hints in documentation. So my first
question is - is what I see is an expected behavior and if yes, how to
come from it to smooth trajectory to reach the goal?

The second question is regarding the bounds of the state space. My map
has dimensions about 1000x1000. So I thought that changing the X and Y
bounds for the state space and setting start and goal to let's say
50,50,0 and 500,800,0 would lead me to similar solution as with
original -1:1 bounds. However, in this case the planner did not found
the path from start to goal. It said that path was found, but the
output definitely shows that the goal is not reached. I would also
appreciate if someone can give some hints where to look at to make
this case running. My guess was that playing with
setMinMaxControlDuration() and setPropagationStepSize() functions
might be useful, but it does not lead to improvements. I have to admit
that I was just guessing parameters, so if there are explanations on
how to select appropriate ones, it would be helpful.


Thank you very much,
Andrey.