High quality desktop video conferencing using agent based systems

This paper describes a new model for building high quality desktop video conferencing systems. The model has a distributed agent based architecture as the underlying framework, where agents cooperate and coordinate efficiently to maintain good quality of service parameters during the video conference session. The model is self-adaptive and permits the system to adjust itself in real time to optimally respond to a set of rapidly changing conditions and parameters. It employs graceful degradation to enable the conference to continue, possibly at a lower pace and perhaps with worse performance, even if some resources (hardware or software) are disabled. Basic learning mechanisms allow the computational unit, an agent, to improve in time as more knowledge about the environment is acquired. A set of performance enhancers are used to compensate hardware overloads and maintain good system throughput when external factors are affecting the environment. The performance enhancers have quality evaluators, a set of internal monitors that analyze the system behavior and activate/deactivate the compensation loop