A study of incremental redundancy hybrid ARQ over Markov channel models derived from experimental data

Abstract

In this paper, we process channel Signal-to-Noise-Ratio time series gathered in the proximity of the Pianosa island, Italy, in Summer 2009. These traces are used to model the performance of capacity-achieving code ensembles as employed in an Incremental Redundancy (IR) Hybrid Automatic Repeat reQuest (HARQ) error control scheme. We apply a code-matched channel state quantization technique aimed at representing channel evolution over time with low quantization error; the evolution of the channel among the quantized states is then represented using a Markov model, over which we base the analytical evaluation of IR-HARQ performance. Results confirm that IR-HARQ consistently improves link performance with respect to Type I HARQ. In addition, we observe that the different channel statistics due to different transmitter and receiver placements, as well as to the acoustic propagation conditions considered in our scenario, have an impact on HARQ performance. This impact is correctly captured by our Markov model, suggesting good adherence of the model to actual channel behaviors. The validation of the models (by simulating over different traces than those used to train the models) suggests that they are robust to moderate non-stationarity, making them good candidates to give a compact representation of the channel behavior, e.g., in network simulators.