Opportunistic scheduling using ARQ feedback in multi-cell downlink

We study cooperative, opportunistic multiuser scheduling using ARQ feedback in multi-cell downlink systems. Adopting the cell breathing ICI control mechanism, we formulate the scheduling problem as an infinite horizon discounted reward partially observable Markov decision process and study two scenarios. When the cooperation between the cells is asymmetric, we show that the optimal scheduling policy has a greedy flavor and is simple to implement. Under symmetric cooperation, we link the scheduling problem with restless multiarmed bandit processes and propose a low complexity index scheduling policy. The proposed index policy is essentially Whittle's index policy, if the scheduling problem is Whittle indexable. Extensive numerical experiments suggest that the proposed policy is near-optimal.