Question about rotated bounding rectangles

Hi all, is it more efficient to use the separate axis theorem for 2D rotated bounding box collision checks or is it better to just rotate the 4 corners of each box, using those 4 corners to make 4 lines and then just do line-line collision checks again each rectangle side? In either case I’ll have outer bounding circles to check they’re in range to reduce the amount of SAT/line-line checks are performed.
Also if anyone can think of any positive/negative points for using either I’d appreciate it. :slight_smile:

Thanks,

Paul

I personally use SAT for Rotated bounding box collisions, it is very fast and accurate and it can be expanded for collision detection between rotated 3D boxes and polygons. Although for 3D, it is commonly better to use an algorithm called GJK (Gilbert–Johnson–Keerthi) which will outperform SAT for most 3D cases.

Thanks for your quick reply, I’ll look it up. :slight_smile: