Delay-Differentiated Scheduling in a Fading Channel

This paper studies scheduling in a time-varying wireless channel where different users may have different delay-requirements. The scheduling problem is formulated as a dynamic programming problem, taking into account both user-specific delay penalties and the time-varying channel state. A scheduling rule similar to the exponential rule is derived, applying dynamic delay-penalized optimization. Threshold-based scheduling optimizes the throughput-delay trade-off in a single-user memoryless channel. Modifications of the optimal single-user scheduling rule to a multiuser channel are studied both in a time-correlated and in a memoryless environment. Numerical experiments suggest that a simplified threshold derived for a memoryless setting performs well in a time-correlated channel. The applicability of the scheduling model to a virtual fading channel with relay nodes is discussed.