A Markov analysis of automatic repeat request for video traffic transmission

This paper presents a study of the automatic repeat request (ARQ) technique applied to the transmission of multimedia traffic, e.g., video content. In the literature, retransmission-based techniques are usually investigated by means of queueing theory and assuming a homogeneous flow of identical packets, which are sent and possibly retransmitted all in the same way. However, multimedia packets are the result of an incremental encoding that leverages spatial and temporal redundancy, which is naturally present in the raw data. As a result, the flow is inherently made of packets with different roles, which should also be treated differently by the ARQ mechanism. Thus, we assume that different levels of error protection are applied, and also we model the decoding process at the receiver as accounting for a dependence relationship among the packets. Moreover, since error correlation has a strong impact on the performance, we consider a transmission over a Markov channel where we tune not only the error probability but also the average error burst size. This enables the derivation of several performance metrics in an entirely analytical manner via Markov analysis. Finally, some numerical results are explored and possible applications on the development of guidelines for multimedia transmission are discussed.