MALDIVES
Measurement and Analysis of Large Distributed VIrtual EnvironmentS
Previous research in DVE mainly based their design on assumptions
on how avatars behave, and use randomly generated movements to
evaluate their design. We believe that there is a gap between these
assumptions and randomly generated movements to how real avatar
behave. To bridge the gap, we collected a large number of avatar
movement traces from Second Life. Our analysis of the traces lead
to several new insights to design of caching, prefetching, zoning,
load balancing in traditional client/server architecture as well
as peer-to-peer DVE.
Cubist
3D Mesh Streaming
3D triangular mesh is becoming an increasingly important data type for
networked applications such as digital museums, online games, and virtual
worlds. A high quality 3D mesh can consists of billions of polygons,
giving a total data size in the order of gigabytes, even after lossless
compression. Disseminating such 3D meshes over a resource-constraint
network for on-demand viewing and rendering presents many technical
challenges. We are investigating how to efficiently and effectively
support streaming of large 3D meshes to large number of clients over the
Internet. This project is partially supported by MOE ARF Grant
232-000-306-112. (September 2007 - September 2009).
DAVIS
Distributed, Automated Video Surveillance
In next generation video surveillance system, video
streams are not watched by humans most of the time, but, instead, are
processed by automated software employing computer vision algorithms.
Computer vision has a much lower requirements on video quality than
human vision. Exploiting this fact, surveillance video streams can be
sent at a low quality without affecting the accuracy of the vision
algorithms.
TCP UREL
Unreliable TCP
We propose an alternative congestion
controlled, unreliable transport protocol based on TCP. Unlike DCCP, our
protocol is merely a TCP option, and therefore can be deployed
immediately, is friendly to TCP, and evolves with TCP. The idea behind
our proposal is simple -- put new data in retransmitted segments. An
implementation on FreeBSD requires only 750 lines of code.
Plasma
A Scripting Language for Manipulating Media Streams
Plasma is a scripting language designed to make it easy to manipulate
live media streams. One can easily compose a new stream from multiple
media sources, including cameras, stored media files, and live streams.
Plasma is event-based. One can bind an action to UI event (e.g. mouse
click), network event (e.g., packet losses) or content (e.g. motion in
video). Plasma is useful to building applications such as video
surveillance and broadcast production.
Dimestr
Multi-source Media Streaming
Distributed media streaming, or multi-source media streaming, exploits
both server and path diversity to improve the quality of the media
streams received. Under this communication model, multiple senders
collaboratively and simultaneously stream a media content to a receiver.
Such model poses breaks some fundamental assumptions in traditional
media streaming systems. We are investigating transport and application
protocols to support this new model, in particular, in error control
and congestion control.
