|
How to distribute virtual reality
Amsterdam, 11 November 97
The University of Antwerp is developing a software platform for building virtual reality applications. Kris Demuynck explained his project during the November Euro-VR Mini-Conference in Amsterdam.
The general idea is to create a flexible and dynamic network system which is easy to extend in order to distribute virtual reality. The main topics of this research are concentrated on the analysis of the performance and on the object-oriented design of the system enabling the researchers to shape building blocks. The virtual environment is composed of entities which have to interact with other entities. This includes a dynamic approach of the entity's internal state and appearance. Typical for the concept of virtual reality is the fact that a human participant is also considered as an entity.
In the implementing phase, processes are being distributed over a heterogeneous Unix network after which they start communicating with each other. This interaction might sometimes result in a network overload. One way to improve the process traffic, is trying to optimise the network by means of creating latency and adapted bandwith. Another solution is to send fewer and smaller messages. Last but not least, possible problems can be eluded by clever techniques. Kris Demuynck is referring here to the use of coherence principles and to the extrapolation of the object's movement. Instead of sending the whole constellation of a changed entity, only the updates are sent.
The type of network which is needed for distributed virtual reality depends on the load balancing. In Antwerp, the research team has started the search for a global optimum which of course remains a theoretical model. Factors that influence the project are the migration of processes as well as the finding of a better host. In any case, the core system is almost finished now and runs on a VR desktop with the additional advantages that the system is flexible and dynamic and delivers a good performance.
The use of multi-cast, allowing less messages, sound collision detection and parsing VRML enabling multi-user VRML, are still under development. If you wish to follow the next steps in the construction of this VE platform, check out RUCA University of Antwerp web site.
Leslie Versweyveld
|