Multi-objective optimization for peer-to-peer multipoint video conferencing using layered video

TLDR: A new peer-to-peer architecture for multipoint video conferencing is presented, based on use of layered video coding with two layers at the end points, which allows each conference participant to see any other participant at any given time under all multipoint configurations of any number of users.

Abstract: We present a multi-objective approach and corresponding formulations for the optimal operation of a peer-to-peer multipoint video conferencing system. The system aims end-points with low bandwidth connections (i.e., single full-quality video in and out) and makes use of layered video to achieve that each participant can view any other participant’s video at anytime. This may cause some of the peers receive lower quality video. Moreover, since the peers may have to forward the video they receive, this may cause larger delays for the peers that receive the video after it is forwarded by several peers. Objective formulations to determine the number of lower quality video receiving peers and the delay experienced by the peers are derived. A multi-objective optimization approach for minimizing both simultaneously is described. An extension that allows multiple video requests from the participants with sufficient bandwidth is proposed. Formulations to minimize the number of lower quality video receivers while maximizing the number of additional video requests are presented. A multi-objective optimization technique assigning importance weights to each of these objectives and its sensitivity to changes in the weights are shown. The use of multi-objective optimization techniques within a system is demonstrated through example scenarios. The effects of our optimization approach on the percentage of base quality receiving peers are examined through simulations.