I got a little problem here, I'm kinda stuck with my diploma thesis. I hope, you can bring me back on track:
In my diploma thesis among other things I wanted to write a collision constraint solver with special regards to large mass ratios, but it seems, that I misjudged the problem in some way. I thought the problem would arise with every collision, because of slow convergence of the lcp solver, but now I realized that only few special configurations cause this issue with sequential solvers. And in addition to that, I develop the solver for a non-gravity scenario, so this problem is just more unlikely to happen at all (no stacking).
In a nutshell, I want to enhance/replace Bullet's built-in SI solver, for a special scenario. The scenario is in space, like simulating a whole star system with celestial bodies, small spacecrafts and everything with real sizes, distances and so forth.
So I have to deal with:
- large distances (already solved)
- large velocities (already solved)
- very huge and very small objects
- large mass ratios
I don't really know, if someone can help me with this at all, but I'm like a little tiny bit desperate already, so I just give it a shot...
So if anyone has maybe some kind of idea what problems with the SI solver could arise in such a scenario of real space simulation and maybe even how one could overcome it, I would love to here something about it, I just don't really know what to do right now...
