
Here are some test cases to demonstrate the algorithm. We implemented the algorithm with C++. Our tests presented below were ran on
a Pentium 4 1GB RAM. Addressing the case in which both boxes are moving is
unnecessary, because we can always change our coordinate system to be one of the
boxes coordinate system, and thus one box is static and the other dynamic. Linear Case
In this case we can see an example of a linear motion. The collision type is cornerface (or in SOR  octantvertex respectively). This example's motion matrix is given by: The equation compatible with the colliding vertex and face is: And the initial collision time is t=0.47. To run the collision prediction algorithm took us, in this case, 0.0128 seconds on a Pentium 4 with 1Gb RAM. For further discussion in this example, please refer to the paper in the Download section. 2nd Degree Case
In this case we can see an example of a 2nd degree motion. The collision type is edgeedge (or in SOR  arcarc). This example's motion matrix is given by: The equation compatible with the colliding edges is: And the initial collision time is t=0.654, and the exit time is t=0.889. To run the collision prediction algorithm took us, in this case, 0.052 seconds on a Pentium 4 with 1Gb RAM. For further discussion in this example, please refer to the paper in the Download section. 4th Degree Case  with no Collision
This example's motion matrix is given by: To run the collision prediction algorithm took us, in this case, 0.103 seconds on a Pentium 4 with 1Gb RAM. For further discussion in this example, please refer to the paper in the Download section. 
