The effects of multiple routing on the end-to-end average distortion

Consider a communication system and a memoryless, loss-tolerant and delay-sensitive source the output of which is transmitted through an arbitraty network. In multiple routing the encoder output is distributed among several independent paths as specified in Fig. 1. The percentage of packets that follows route "i" is denoted by q_i where Σ_i=1^K qi = 1. The total delay experienced by a packet on its journey to the destination is a function of several parameters such as network congestion, route quality, link capacity etc. If the source data is delay-sensitive (e.g., voice or video) then these network parameters have a great impact on the achievable distortion at the receiver. The performance measure that is of interest is the average end-to-end distortion under various network loading. The purpose of this paper is to show that intelligent multiplexing (i.e., smart choices for q_i i = 1,2,...,K) plays an important role in reducing the achieved average end-to-end distortion at high network loading.