This proposed research will design, implement and test algorithms to automate the production of webcasts and distributed collaborations using Internet multicast streaming media technologies. We have considerable experience producing webcasts for seminars, workshops, special events, and class lectures. We have also produced videoconferences using Internet Mbone and Access Grid (AG) technologies. Our experience indicates that the cost of setting up and operating the required facilities limits the use of this technology.

Our long-term goal is to improve the webcast and videoconference experience and to reduce the cost of production. Production cost can be reduced by continuously improving the physical facilities and equipment to exploit new technologies and by replacing expensive human labor with software automation. Over the past several years we have developed a software architecture and collection of tools to automate the production of webcasts including: 1) a broadcast manager tool for launching standard configurations of processes for an event, 2) a director's console tool for manually controlling event streams (e.g., camera control, source video switching, audio levels, etc.), 3) a question monitor tool to support seamless interaction with remote participants, and 4) a transcoding gateway tool for managing multiple transmissions using different streaming technologies, media formats and bit rates. Now, we need to reduce the human labor required to produce events.

We have experimented with a prototype virtual director tool that automated the decisions made by a human producer/director. The prototype used rules to express behaviors that the event should satisfy. The rules involved decisions to switch camera sources and position cameras to participants asking questions. This prototype did not work as well as it might have worked when compared with events produced by professionals primarily because of limitations in the heuristics that could be specified and the behaviors that could be implemented. Most notably, the system did not allow the introduction of new services as independent processes that could signal events to the virtual director. The proposed research will re-implement this prototype using a better software architecture. This implementation will fix the problems with the prototype and allow us to experiment with more sophisticated services and production heuristics. A key element of the proposed research is to develop a scripting language for specifying a webcast or videoconference production.

The AG videoconferencing model is an emerging standard for high-bandwidth distributed collaboration on the Internet. Current production models require one or more producer/directors at each site to control local a/v equipment and manage display of remote streams. The proposed research will incorporate support for controlling window placement, local camera control, and source audio/video transmissions for interaction with remote participants in an AG videoconference.