Evaluating the network and usability characteristics of virtual reality conferencing

This paper presents the results of the BT-funded 'Inhabiting the Web' project which conducted a series of twenty virtual meetings using the MASSIVE virtual reality teleconferencing system between five universities and BT Laboratories. The aims of the project were, firstly, to assess the characteristics of the network traffic generated by MASSIVE and therefore, by extension, the network traffic generated by other peer-to-peer unicast-based collaborative virtual environments, and secondly, to identify key human factors issues arising from the long-term and regular use of such a system. The paper gives an overview of the MASSIVE system and the organisation of the meetings and then presents an analysis of its results. The network analysis involves a statistical analysis of system logs in order to build a profile of typical user behaviour. This profile indicates how much users might be expected to engage in basic actions such as moving and speaking during a typical meeting and also to what extent these actions might be correlated among users. When combined with ail understanding of MASSIVE's network protocols, this leads to a model to predict the network traffic generated by varying numbers of users. It also suggests new design possibilities such as 'group portals', caching of worlds and the use of multicast protocols. Usability reflections address the issues of embodiment, navigation and use of the spatial model of interaction as implemented in MASSIVE. New design possibilities are identified including higher level support for gestures oil embodiments and so-called object-centred and context-sensitive styles of navigation.