You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It looks like this line if (curIV[0] && curIV[1] - curIV[0] < tTol) { is avoiding finding a root if the start of the interval equals 0, right?
By default, our CCD will converge to any solution as long as the interval width is less than the tolerance (even if it contains 0). There is a CMake option TIGHT_INCLUSION_WITH_NO_ZERO_TOI which will continue refining if the time of impact is zero. Can you try with that ON?
It would be great if it can report the ground truth "no collision", as this could still limit the step size taken about 1/50 smaller. Maybe figuring out a way to set more adaptive tolerance can help? Anyway this is just my curiosity.
For the PT CCD case below, TICCD returned 0 TOI.
The format is
I used the TICCD from the CCDWrapper repo (latest commit), and I called it like
The output is
However I implemented a simpler version of TICCD according to the paper as
and it kinds of worked on this case and output
which with tighter convergence tolerance might be able to reach the ground truth solution -- no collision.
In my implementation, I didn't include the epsilons in formula (5) in the paper. Is it the main cause here?
The text was updated successfully, but these errors were encountered: