A system for automatic modeling of anatomical joint motion for use in the Virtual Reality Dynamic Anatomic (VRDA) tool is described. The modeling method described in this article relies on collision detection. An original incremental algorithm use this information to achieve stable positions and orientations of the tibia on the femur for each angle considered between these two components on the range of motion. The stable states then become the basis for a look-up table employed in the animation of the motion of the joint. The strength of the method lies in its robustness to animate any “normal” anatomical joint, given a set of kinematic constraints for the joint type as well as an accurate 3D geometric model of the joint. The demonstration could be patient specific (based on a person’s real anatomical data from an imaging procedure such as CT scanning) or scaled from a generic joint based on external patient measurements. The modeling method has been implemented on a generic knee model for use in the VRDA tool.