Achievable rates and rate selection algorithms for incremental redundancy (IR) hybrid ARQ (HARQ) wireless systems
In this thesis, we first characterize the achievable rate region and outage probability for generalized time × bandwidth adaptive IR HARQ systems in block fading. We then obtain throughput expressions for non-adaptive and adaptive IR HARQ systems given delayed channel mutual information feedback using renewal process theory. We derive the optimal HARQ rate selection algorithms for conditional throughput under a conditional outage probability constraint in both scenarios. ^ For the adaptive IR case, we formulate the N-stage rate selection problem as a Markov Decision Process over a finite horizon; since the optimal algorithm fast becomes intractable as N grows, we use the fact that a good approximation of the cost function is minimized at a post HARQ outage equal to the desired target to derive simpler variants that compute a uniform target conditional outage for the penultimate stage states. Independently, by restricting the outage constraint to be satisfied at every stage, we derive a relatively low complexity Dynamic Programming algorithm and study an easing of this restriction to improve this solution by combining it iteratively with the aforesaid concept of penultimate stage uniform target conditional outage. ^ We numerically evaluate throughput/outage performance for these and more algorithms under time correlated single path Rayleigh fading and compare with commonly employed other rate adaptation algorithms, characterize HARQ gain components of time diversity and early termination/adaptivity and study the impact of feedback and opportunistic scheduling at various Doppler. ^
Saul B. Gelfand, Purdue University.
Engineering, Electronics and Electrical