On Delay-Power Tradeoff of Rate Adaptive Wireless Communications with Random Arrivals

In this paper, we study delay optimal scheduling of bursty data traffics over multi-state time-varying wireless channels, where bursty packet arrival in the network layer, queueing behavior in the data link layer, and rate adaptive transmission with flexible modulation in the physical layer are jointly considered from a cross-layer perspective. To achieve a minimum queueing delay under a power constraint, a probabilistic queue-aware and channel- aware cross-layer scheduling policy is proposed, and characterized by a Markov chain model, where the transmission rate, i.e., the number of packets delivered in each slot, is selected with probabilities based on the buffer and channel states in this slot. To reveal the optimal delay-power tradeoff, we formulate a non-linear optimization problem, which, however, is very challenging to solve. To make it tractable, we convert the optimization problem equivalently into a Linear Programming (LP) problem, which helps us achieve the optimal three-dimensional threshold-based scheduling policy analytically. It is found that the source should select one transmission rate jointly based on the channel state and the backlog in the queue.