Integration of collision and graphics layers at sub-entity level
Motivation
One of the most astonishing features shown in Crysis is the tight integration of the collision and graphics layers. This can be seen in the way plants bend when the character passes through them, or the sensation of a completely interactive world where everything can be touched, bent, broken, shot or utterly destroyed. There is a sense that any level of world subdivision (tree, branch, leaf) that is convenient for collision/physics modeling can be mapped into a convenient graphics structure realistically showing the results of that simulation. The implementation of how collision is mapped into renderable graphics in most current game technology makes it almost impossible to think of achieving such close interdependence between both layers, due to both performance and most importantly, the wrong design in the interface of both systems: the collision layer is a black box accessible only as one-piece entities, no matter how many geometries it contains; and typical scene graphs lack both the capability to dynamically tweak the inner hierarchical structure of a pre-defined graphical construct, except internally (purely graphic animation). This is why ragdoll-character integration is so difficult to achieve, and why something that looks as simple and obvious as colliding particles looks as such an advanced feature.
This project tries to prove wrong the idea that graphics and collision layers are only capable of interaction at the entity (global transform) level. I will develop a basic framework for both collision/physics simulation and graphics that allows both a physics-controlled structure arbitrarily map itself correctly into its graphics representation. This interaction can be applied at several levels:
· Map rigid-body physics to a rigid, hierarchical graphic transform
· Map hierarchical, joint-based collision structures to deformable, skinned meshes
· Map arbitrary damage or destruction effects into its graphic visualization: bending, bumping, crushing, breaking or plain blowing the object to smithereens.
· Map arbitrary deformation of a solid mesh into a graphic mesh at vertex level (eg. cloth animation)
· Map arbitrary fluid deformation and dynamics into a graphic fluid volume
With this project I plan to devise an alternate representation for hierarchical graphics, and a flexible interface for modifying the internals of such a structure, as well as the internals of defining a collision layer.
