Middleware services for distributed virtual environments

Abstract

Distributed Virtual Environments (DVEs) are virtual environments which allow dispersed users to interact with each other and the virtual world through the underlying network. Scalability is a major challenge in building a successful DVE, which is directly affected by the volume of message exchange. Different techniques have been deployed to reduce the volume of message exchange in order to support large numbers of simultaneous participants in a DVE. Interest management is a popular technique for filtering unnecessary message exchange between users. The rationale behind interest management is to resolve the "interests" of users and decide whether messages should be exchanged between them. There are three basic interest management approaches: region-based, aura-based and hybrid approaches. However, if the time taken for an interest management approach to determine interests is greater than the duration of the interaction, it is not possible to guarantee interactions will occur correctly or at all. This is termed the Missed Interaction Problem, which all existing interest management approaches are susceptible to. This thesis provides a new aura-based interest management approach, termed Predictive Interest management (PIM), to alleviate the missed interaction problem. PIM uses an enlarged aura to detect potential aura-intersections and iii initiate message exchange. It utilises variable message exchange frequencies, proportional to the intersection degree of the objects' expanded auras, to restrict bandwidth usage. This thesis provides an experimental system, the PIM system, which couples predictive interest management with the de-centralised server communication model. It utilises the Common Object Request Broker Architecture (CORBA) middleware standard to provide an interoperable middleware for DVEs. Experimental results are provided to demonstrate that PIM provides a scalable interest management approach which alleviates the missed interaction problem.