Hi everyone,
Has anyone come across any good techniques for performing real-time proximity queries on the GPU beyond interference detection? I'm talking about things like contact features and penetration depth. Most of the academic literature out there is a bit limited in this area. It seems like most algorithms just perform interference detection on the GPU then revert back to software techniques to get more useful data.
I know of a few examples from Hoff and co. which perform 3D proximity queries by creating Voronoi diagrams using the GPU, but as far as I know these data structures are moved back into software before they can be queried. The readbacks make the algorithm unsuitable for real time.
So for a rigid body simulation performed mostly on the GPU, how would one go about calculating penetration depth? For simplicity, lets say we're using convex objects only. Any suggestions?