Channel-Adaptive Schedulers with State-of-the-Art Channel Predictors

In this paper we study the effects of using state-of-the-art channel predictors with two channel-aware schedulers: a channel-aware round robin, and the proportional fair scheduler. We evaluate the performance for the case when the scheduler queues are always full and for the case when this does not happens. The round robin scheduler proves to be only marginally influenced by the inaccurate predictor in both cases. The proportional fair scheduler, however, can become very unfair when few wireless terminals have data to be scheduled, some of them have a very good channel and the channel prediction is slightly inaccurate. When the queues are not full, however, this behaviour is no longer a problem. But, in this case, there is not much to win by the use of a complex scheduler and complex channel predictor; a simple channel-aware round robin and a one-step predictor achieve similar performance.