Performance evaluation of client-side video stream quality selection using autonomous avatars

Abstract

In this paper we present test results performed on our framework for networked virtual environment (NVE) applications that incorporates real-time video communication between avatars. The primary goal of the architecture is to provide efficient scalability for large scale networked virtual environments. To realize this, our solution maximizes client responsibilities and relies on direct client-to-client communication streams. By employing multiple multicast groups to channel the video streams, we achieve bandwidth adaptation at client side with minimal server intervention. This results in a reduced server-load and at the same time guarantees a highly scalable end-result, depending only on the available processing power of the individual connected clients.