Level of Motion Detail in Virtual Reality

Roman Berka
Dept. of Computer Science and Engineering
Czech Technical University
Karlovo nam. 13
121 35 Prague


Our paper presents an optimization technique which
modulates the computation complexity of the animation in virtual
reality (VR). As this technique is based on approaches commonly used
in the domain of LOD, we call it LOmD (Level of motion Detail). We
extend the classical LOD mechanism based on geometry removal according
to observer's distance from a target object, velocity of the model, and
position of the model in the periphery of the visual field. The LOmD
technique is based on the idea that our visual system is not able to
recognize differences in the quality of motions produced by different
animation techniques (e.g. simple linear interpolation vs. physically
based simulation) under different visual conditions (e.g. distance or
eccentricity). Varying the motion detail we can balance the workload
of VR system in such a way that the only motions important for the
user's view are refined using more complex simulation whereas other
motions are linearly interpolated to quarantee the motion
continuity. In order to achieve automatic behavior of LOmD mechanism
we designed the visual acuity model based on visual perception which
links the viewing conditions to the appropriate level of motion
detail. The proposed method has been tested on implementation of a
simple VR model. The results are presented at the section 5 of
the paper.