There are many such algorithms, and it is an active area of research especially in the area of soft body physics (cloth, human tissue, etc) where precise collis

There are many such algorithms, and it is an active area of research especially in the area of soft body physics (cloth, human tissue, etc) where precise collision handling is very important. Most rigid body simulations use discrete collision detection with fixed time steps, so it only checks for collisions at specific points in time. This is fast, but can cause issues where fast moving objects passing through thin objects (such as a bullet through a pane of glass). The alternative is called Continuous Collision Detection in which you calculate exactly when two objects will collide (or have collided) and can then simulate up to that point (or roll back to that point) and deal with the collision. One naive technique to do this is after integrating your simulation forwards, you use the volume each object swept out in that time step as its collision geometry instead of just the shape of the object at its new position. This solution has issues; it doesn't tell you when two objects collided and it can give false positives, but it works ok for rigid body simulations. Some recent papers on the subject with more exact (and complicated) techniques: http://www.cs.columbia.edu/cg/ACM/ http://www.cs.columbia.edu/cg/ACM/ http://www.cs.ubc.ca/~rbridson/docs/brochu-siggraph2012-ccd.pdf http://www.cs.ubc.ca/~rbridson/docs/brochu-siggraph2012-ccd....