In the OwenStateSpace, the length computation can fail if the numerical search fails, as mentioned in:
|
// Discontinuities in the Dubins distance and, by extension, radiusFun can cause bracket_and_solve_root to fail. |
|
if (std::abs(radiusFun(radius)) > 1e-5) |
|
return {}; |
I noticed this actually fails quite often, and probably influences the quality of the path found by the planner.
The failure rate is almost consistently 25% given 100,000 randomly sampled states.
[benchmark_dubins_airplane-1] Failure rate: 24922 / 100000
@mamoll FYI, not sure if you would have some insights on how to make this more robust
In the OwenStateSpace, the length computation can fail if the numerical search fails, as mentioned in:
ompl/src/ompl/base/spaces/src/OwenStateSpace.cpp
Lines 128 to 130 in 27773ae
I noticed this actually fails quite often, and probably influences the quality of the path found by the planner.
The failure rate is almost consistently 25% given 100,000 randomly sampled states.
@mamoll FYI, not sure if you would have some insights on how to make this more robust