A new method for formulation, solution, and sensitivity analysis of collision detection of convex objects in motion is presented. The collision detection problem is formulated as a parametric programming problem governed by the changes in the relative translation and relative rotation between the two objects considered. The two parameters together determine all the possible relative configurations between two moving convex objects. Therefore, solving this parametric problem allows for knowing the proximity information for all the possible configurations of the objects. We develop a two-step decomposition procedure to solve this parametric programming problem, and show that the solution is a convex function of the two parameters. This convexity feature enables an archive of the proximity information and sensitivity analysis for the collision detection problem.