An Adaptive Hybrid ARQ Protocol for Continuous-Time Markovian Channels

This paper analyzes the throughput of an adaptive hybrid ARQ communications protocol used with a continuous-time two-state Markovian channel. The proposed protocol adapts to changes in the channel state. i.e., changes in noise level, by selecting one of two block error-correction codes for each packet according to an estimate of the next channel state; the estimate, based on the error pattern in the previous packet, is fed back to the transmitter via the acknowledgment. We compare throughputs and computational costs of three next-state estimators. The first, an optimal estimator, is shown to be impractical. A second, suboptimal, estimator is found to require negligible computation and memory. An improved suboptimal estimator is also considered. Throughputs are evaluated; numerical results are presented. The throughput improvement due to adaptation is shown to be greater for the continuous-time channel than for the discrete-time channel with sirmilar parameters. The improvemient is also less sensitive to the parameters.