Collaborative rendering over peer-to-peer networks

Abstract

Physically-based high-fidelity rendering pervades areas like engineering, architecture, archaeology and defence, amongs others. The computationally intensive algorithms required for such visualisation benefit greatly from added computational resources when exploiting parallelism. In scenarios where multiple users roam around the same virtual scene, and possibly interact with one another, complex visualisation of phenomena like global illumination are traditionally computed and duplicated at each and every client, or centralised and computed at a single very powerful server. In this paper, we introduce the concept of collaborative high-fidelity rendering over peer-to-peer networks, which aims to reduce redundant computation via collaboration in an environment where client machines are volatile and may join or leave the network at any time.