Opportunistic scheduling with worst case delay guarantees in single and multi-hop networks

We first consider a multi-user, single-hop wireless network with arbitrarily varying (and possibly non-ergodic) arrivals and channels. We design an opportunistic scheduling algorithm that guarantees all sessions have a bounded worst case delay. The algorithm has no knowledge of the future, but yields throughput-utility that is close to (or better than) that of a T-slot lookahead policy that makes “ideal” decisions based on perfect knowledge up to T slots into the future. We then extend the algorithm to treat worst case delay guarantees in multi-hop networks. Our analysis uses a sample-path version of Lyapunov optimization together with a novel virtual queue structure.