In a distributed walkthrough system, a geometry database server maintains a huge collection of virtual objects and their positions in the virtual environment. Human viewers utilizing database clients connected to the server through networks may view the objects by walking through the virtual environment. To model a large and complex environment, the database is often very large.

Several approaches have been proposed to distribute objects from the server to the clients in distributed virtual reality applications. Most existing systems, such as DIVE, SIMNET, and VLNET, transmit the whole geometry database to the clients before the walkthrough starts. To use the systems in an Internet environment, either the size of the geometry database needs to be small or the viewer has to wait for a long pre-loading time before starting the walkthrough. Another approach to distribute the objects in the geometry database is to send them on demand to the clients during the walkthrough. This approach employs a standard client-server architecture, in which a central server maintains the geometry database of the virtual environment and distributes data to clients when requested. Both approaches, however, require the models of the visible objects to be completely transmitted before the corresponding images can be rendered. This might result in uneven frame rates in a congested network such as the Internet and will cause a jittered walkthrough. It also limits the scene complexity that the system can handle.

In this project, we are developing a distributed walkthrough system for use in an internet environment. This system has two major features:

• On-demand transmission: The system only transmits the region of the environment visible to the client to its machine. We define the user scope and the object. The user scope indicates how far a user can see and the object scope indicates how far an object can be seen. We only need to download those objects whose scopes overlap with the user scope to the client machine.
• Progressive transmission and reconstruction: Sending the whole model of each visible object to the client machine on-demand may cause a possibly long pause in the walkthrough. Instead, our system encodes each model in the form of a progressive mesh for transmission. In most situations, the visibility of each object only changes slightly from one frame to the next. Hence, we only need to send a small portion of the model to the client machine, where the object model is progressively reconstructed. If the network is too busy, we may still use a low resolution model of the object for rendering.

Click here for an animation sequence produced by our prototype system. (The file is gzip'ed. It is very large. Compressed size is 4.6MB and uncompressed size is 17MB.)

Publications:

• Beatrice Ng, Rynson Lau, Antonio Si, and Frederick Li, "Multi-Server Support for Large Scale Distributed Virtual Environments," To appear in IEEE Transcations on Multimedia.
• Jimmy Chim, Rynson Lau, H.V. Leong, and Antonio Si, "CyberWalk: A Web-based Distributed Virtual Walkthrough Environment," IEEE Transactions on Multimedia, 5(4):503-515, Dec. 2003.
• Rynson Lau, Beatrice Ng, Antonio Si, and Frederick Li, "Adaptive Partitioning for Multi-Server Distributed Virtual Environments," Proceedings of ACM Multimedia, pp. 271-274, Dec. 2002.
• Beatrice Ng, Antonio Si, Rynson Lau, and Frederick Li, "A Multi-Server Architecture for Distributed Virtual Walkthrough," Proceedings of ACM VRST, pp. 163-170, Nov. 2002.
• Danny To, Rynson Lau, and Mark Green, "A Method for Progressive and Selective Transmission of Multi-Resolution Models," Proceedings of ACM VRST, pp. 88-95, December 1999.
• Jimmy Chim, Rynson Lau, Antonio Si, H.V. Leong, Danny To, Mark Green, and M.L. Lam, "Multi-Resolution Model Transmission in Distributed Virtual Environments," Proceedings of ACM VRST, pp. 25-34, November 1998.
• Jimmy Chim, Mark Green, Rynson Lau, H.V. Leong, and Antonio Si, "On Caching and Prefetching of Virtual Objects in Distributed Virtual Environments," Proceedings of ACM Multimedia, pp. 171-180, September 1998.
• Rynson Lau, Mark Green, Danny To, and Janis Wong, "Real-Time Continuous Multi-Resolution Method for Models of Arbitrary Topology," Presence: Teleoperators and Virtual Environments, MIT Press, 7(1), pp. 22-35, February 1998.
• Rynson Lau, Danny To, and Mark Green, "Adaptive Multi-Resolution Modeling Technique Based on Viewing and Animation Parameters," Proceedings of IEEE VRAIS, pp. 20-27, March 1997.
• Veysi Isler, Rynson Lau, and Mark Green, "Real-time Multi-resolution Modeling for Complex Virtual Environments," Proceedings of ACM VRST, pp. 11-20, July 1996.

Last updated on 9th October, 2004.