Virtual environments include static scenes and moving objects. These mobile objects are either manipulated by the user or animated by a behavioural simulation following the rules of physics or the interpretation of animation scripts. For a “realistic’’ multimodal rendering, mainly in an interactive mode, the behaviour of the simulation and hence that of the objects has to be credible for the operator’s perception. The possible interpenetration of virtual objects, when they collide and when they interact, reduces this realism. Moreover, in case of rigid objects, it was observed that non-penetration increases the perception of rigidity which helps offset the limitations of certain haptic interfaces in terms of their capacity of force rendering. However, when we know how a physical simulation engine operates and the foundations of the current graphic libraries, we can very well guess that nothing can be predefined to prevent virtual objects from interpenetrating. To avoid such unrealistic situations, the physical engine controlling the simulation must check at very short and regular intervals whether two objects are occupying the same position in space at the same time. This management of interactions between objects is generally known as collision detection.