This addition the triangle tests performed the static geometry array

any of these meshes, the triangle data that the collision system has in its static arrays will be out of date and will no longer represent what the scene actually looks like to the player. Keep in mind that our collision and render geometry are separate things. So a door that opens for example, would visually appear to present a new path for the player to explore, but if it was registered with the collision system in an initially closed state, it would remain so and prevent the player from crossing the threshold (and vice versa if the door was registered in an open state).

Clearly, we need some way of letting the collision system know that some of its geometry is dynamic. As such, a new mechanism will have to be added to our collision system for such objects. Processing dynamic scene objects will place an even greater burden on our CPU (more so than testing collisions against our static geometry arrays), so we should only use the dynamic object system for meshes or hierarchies that we intend to animate. Geometry that will never change should still be added to the collision system as before for storage in the static arrays. What we will add is a secondary sub-system that will maintain a list of dynamic objects which the collision system needs to be kept informed about. This way it can track positional and/or rotational changes.

It would seem that we have to sweep the sphere along its velocity vector and the dynamic object along its velocity vector and then determine the intersection of the two swept volumes. Indeed that is one possible approach, but it is made complicated by the fact that the shape of the dynamic object might be something as simple as a cube or as complex as an automobile mesh. We know that we can sweep a sphere/ellipsoid easily (which is why it was used to bound our moving entity), but we hesitate to use crude approximations like this when sweeping the dynamic scene objects since we are after exact triangle intersections, not just the intersections of two swept bounding volumes.

143