[ompl-users] Do LTL propositions need to be continuous and/or convex?

[Helio P. F. <xpe...@gm...>]

Hi all,

I started looking into LTL motion planning following Pouria's question.

I see that the demo
<https://ompl.kavrakilab.org/LTLWithTriangulation_8cpp_source.html> uses
convex polygons for propositions, which is certainly the most
cost-effective approach.

But I was wondering, would OMPL's LTL planner still work if a suitable
implementation of non-convex propositions were supplied?

Moreover, what if propositions were discontinuous?

For example, in my system I use a distance map to quickly determine the
distance between a 2D point and the closest obstacle. Let's say I created a
subclass of ompl::control::PropositionalDecomposition
<https://ompl.kavrakilab.org/classompl_1_1control_1_1PropositionalDecomposition.html>
with
an overloaded worldAtRegion(int rid)
<https://ompl.kavrakilab.org/classompl_1_1control_1_1PropositionalDecomposition.html#aab6a90b0083de432493948f8ce6d01fe>
function that set the truth value of an "obstructed" proposition depending
on whether the centroid of the decomposition region rid is closer than a
threshold d to an obstacle, and set the safety automaton to always enforce
that proposition to be false.

Would the LTL planner be able to handle that?

I looked into the source code and didn't see anything that made me think it
wouldn't work, but I would like to hear a second opinion.

-- 
Best regards,
Helio Perroni Filho
E-mail: ma...@xp...
Web: http://xperroni.me
<http://machineawakening.blogspot.com/>