Demand-Driven Geometry Transmission


A network protocol for efficient distribution of geometry data
Project duration: 1995-1997
Contact: Dieter Schmalstieg

Description

The Demand-Driven Geometry Transmission Protocol was designed to distribute geometric models on the fly in distributed virtual environments based on a client-server architecture.

Application

All multi-user virtual environments on the Internet benefit from this technology.

Problems

In highly interactive VR applications, the necessary geometric models must be transported over the network while the system is running. The data must become available before it is needed, which is often defeated by a slow network.

Approach

The fundamental idea is to organize the world into objects that are downloaded on demand. Users invoke a client program that displays the virtual world from the user's point of view and request the required geometric objects in a timely fashion from the geometry database server. The client only maintains in memory what can be seen by the user, which is a relatively small subset of the large world database. As the user travels through the virtual world, this subset is updated incrementally. Efficiency is further improved by managing individual levels of detail rather than complete objects, which reduces memory and network strain.

Publication

Demand-Driven Geometry Transmission for Distributed Virtual Environments :
D. Schmalstieg, M. Gervautz

Some technical background

A distributed geometry database: A server stores geographically dispersed objects. Each client's view is limited to an area of interest (AOI). If AOIs overlap, clients can see each other.
The unit of transmission is not a whole objects, but a single level of detail for one object. The image shows a hierarchical model of a car, composed of a control structure, the "trunk" (shown in yellow). The actual geometry is mounted to the trunk, making only transmission of the necessary pieces of geometry possible. An even more fine-grained control over data transfer is made possible by the integration of another invention: Smooth Levels of Detail.


This page is maintained by Dieter Schmalstieg. It was last updated on October 29, 1997.
If you have any comments, please send a message to dieter@cg.tuwien.ac.at.