I've messed around a bit with the Bullet physics library and was wondering if anyone knows how it works under the hood. It's not brute forcing the collisions, because it can handle thousands of objects.
Does anyone know how they do the approximation? There is a fudge factor you can set for say a 10% boundary around an object, where it's almost like it moves all of the objects and then takes a best guess where they should be placed for the next physics frame. But I'm curious what their approach is called, if it's a general solution.
Look into quadtrees, that's a good way to do collision detection in 2d. Or just partition your world into a fixed grid and only consider collisions with objects in the same grid square. Quadtrees are just a smarter way of doing that.
Does anyone know how they do the approximation? There is a fudge factor you can set for say a 10% boundary around an object, where it's almost like it moves all of the objects and then takes a best guess where they should be placed for the next physics frame. But I'm curious what their approach is called, if it's a general solution.