Wireless Communications With Hard Delay Constraints: Cross-Layer Scheduling With Its Performance Analysis

Hard-delay constrained wireless communication has attracted considerable recent attention because it holds the promise of playing a vital role in wireless Internet of Things (IoT) systems for supporting time-sensitive tasks. Hard-delay constrained communication refers to the case in which the delay-violation probability of packets is zero with a given deadline. How to strike the optimal power-latency-throughput tradeoff with a hard-delay constraint remains open. In this paper, we aim to characterize the hard-delay constrained capacity with a deadline which covers one or multiple coherence time of the fading channel. Specifically, we first conceive a low-complexity and yet sub-optimal cross-layer scheduling scheme to obtain an analytical bound of the hard-delay constrained capacity. Saddlepoint approximation is adopted to obtain its hard-delay constrained throughput as a function of the tolerated delay and average power, which provides us with properties of the achievable bound of the hard-delay constrained capacity. Further, we investigate the optimal joint channel and queue-aware scheduling to derive the hard-delay constrained capacity. In particular, by quantifying channel and queue states and recalling the optimality of threshold-based policies, we approximate the optimal hard-delay constrained scheduling by a deterministic policy obtained based on the quantized Constrained Markov Decision Process. More structures of this optimal scheduling policy are revealed through curve fitting by dividing system states into distinct regions. Finally, the hard-delay constrained capacity is presented as a function of both the average power and maximum delay, with the assistance of both theoretical analysis from the achievable bound and curve fitting.