Learning Methods for CDF Scheduling in Multiuser Heterogeneous Systems

In modern heterogeneous wireless networks, the task of supporting fairness along with user priorities and concurrently achieving the highest possible system throughput is desirable and challenging. In this work, two classes of practical cumulative distribution function (CDF) based scheduling algorithms are developed to achieve these goals. These algorithms are shown to frequently outperform, and are potential alternatives to, the well known Proportional Fair (PF) scheduling method. The first class of algorithms, Nonparametric CDF based scheduling (NPCS) algorithms, are used when the channel fading model is unknown. Herein, the mapping from channel quality information (CQI) to the real CDF is unknown but is constructed exploiting the order statistics of the CQI sequence. The constructed CDF mapping methods are shown to converge to the actual CDF. When the channel model is known, a class of Parametric CDF based scheduling (PCS) algorithms are developed which learn parameters of the channel statistics for the scheduler to use. In our experiments, this Bayesian learning approach results in better system throughput than the NPCS approach. We also show collecting a moderate number of CQI data is enough to achieve nearly the performance of CDF based scheduling with known channel distribution. Throughout the work, CDF based scheduling algorithms are supported by simulations which show that they can effectively support not only fairness but also user priorities and often outperform PF in terms of system throughput.