Cross-Layer Adaptive ARQ and Modulation Tradeoffs

Abstract

We investigate the tradeoff between uncoded adaptive modulation (AM) in the physical layer, and automatic repeat re-quest (ARQ) in the data link layer, on the system throughput. Most of the applicable research has assumed fixed maximum-number-of-retransmissions (MNR); however, we show that applying adaptive MNR (referred to as adaptive ARQ in this paper) based on channel condition, enhances the throughput performance of the system. In the second part of this paper the effect of finite length queuing, AM, and ARQ on system packet loss is studied. As today's networks and applications generate the traffics that are bursty and self-similar, using Poisson processes to model the packet arrivals proves to be an error-prone assumption. Instead, we will apply generalized exponential (GE) and Pareto distributions, which allow for the modeling of a wide range of bursty distributions, to model the packet arrivals. Closed form expressions for the packet loss and the packet dropping probabilities are derived as functions of MNR and transmission rate. We show that employing adaptive ARQ decreases the packet loss rate while improving the throughput.