Packet Latency for Type-II Hybrid ARQ Transmissions over a Correlated Error Channel

Type-II Hybrid-ARQ (Type-II HARQ) has been shown, under certain circumstances, to increase the efficiency and reduce packet loss of data transmissions over a wireless channel. We present a model for evaluating the performance of packet transmissions over a wireless link employing Type-II HARQ error mitigation. Considering a channel subject to correlated bit errors, numerical results for a wide range of channel statistics are presented, illustrating the effect of bit error correlation, bit error rate, and block size on packet latency. 1. INTRODUCTION In this paper we present a numerical model for the design and performance evaluation of wireless packet data networks that employ Type-II Hybrid-ARQ (Type-II HARQ) [14]. A specific application of the model is to support the design and evaluation of end-to-end transport layer flow control algorithms. Due to the inherent variability of a wireless link, the design and evaluation of end-to-end transport layer flow control protocols is a non-trivial task. Ultimately the underlying physical and link layers will determine the statistics of the packet transmission time as seen by transport layer flow control mechanisms. The statistical characteristics of the packet transmission time is of great importance, as it directly determines how transport layer algorithms should set time out timers, and control packet flow rate. Other applications of the model include base station router buffer dimensioning, and determining quality of service constraints. Previous work [1,2,3,7] has provided bounds on the throughput of Type-II HARQ assuming that channel errors are independently and identically distributed. Some work has examined the performance of ARQ protocols under correlated errors [10,11], however these models have been limited to modelling the correlation in successful block transmissions, as opposed to explicitly modelling the correlation of bit errors. For simple ARQ schemes that fix the coding rate for all blocks, the two techniques are equivalent, and it is sufficient to model the correlation in successful block transmissions for performance analyses. However, in order to analyse the performance of ARQ schemes that alter the coderate with each successive block transmission (such as type-II HARQ), it is necessary to explicitly account for correlations in bit errors. One of the contributions of this paper is to model type-II HARQ explicitly, taking into account the correlation in bit errors. Futhermore, little is understood about how correlated errors affect the statistics of the packet transmission time. This