In the design phase, a user of a radiosity system needs very fast updates of the solution after any scene modifications (e.g., adding/deleting/moving objects, changing light source emission characteristics, changing a surface's reflectance properties). In their initial form, radiosity systems only accommodate static environments, and would force a complete re-evaluation of the global lighting situation for each modification. Such modifications are essential for any number of applications (e.g., high-quality animation or Virtual Reality applications), but the cost of repeated radiosity re-computation is not an acceptable one. Extensions to the radiosity approach have since alleviated this problem: Besides methods for pre-defined animation paths and raytracing-based approaches, the idea of exploiting coherences in progressive-refinement-based radiosity methods has been developed since 1990, and recently, with the line-space-hierachy an approach based on the hierarchical radiosity method has been proposed. But still, it is currently impossible to achieve interactive radiosity update rates for moderately complex scenes in which geometry or materials change, so further improvements have to be found.
In this work package, we will emphasize concrete solutions which will allow interactive dynamic updates of lighting information, permitting the use of radiosity solutions for interactive planning and design, in scenes where object geometry and material properties change. We will first present the data structures and algorithms necessary to achieve fast updates of important lighting information such as shadows, which combine the advantages of adaptive, hierarchical, and clustering radiosity methods. Then we will develop an approach which, while potentially less accurate, will allow real-time feedback. We will also develop approaches which will limit temporal aliasing artifacts (flickering) in animations.
The results of this work package are (or will be) presented in the deliverables.