A Predictive Scheduler for FDD Systems Based on the Windowed Empirical CDF

Abstract

Channel aware schedulers represent a convenient solution for the resource allocation problem in wireless systems by leveraging on fast channel fluctuations to achieve a convenient trade-off between sum-rate and fairness. Even though the current decision is conventionally performed with respect to the past assignments and measurements, interest has recently been focused on predictive schedulers to account on the future (i.e., predicted) channel values for the optimization of the current decisions. This paper provides the implementation for a class of predictive schedulers called P-CBS (predictive CDF based scheduler) as they are based on the empirical cumulative density function of the channel quality. Differently from related works on this topic, here it is assumed that the knowledge about the channel quality is achieved through feedback channels from the users as currently envisaged for Beyond 3G standards. The proposed optimal scheduling approach takes advantage of the estimated CDF from available past values in a predictive way. A sub- optimal (and somewhat intuitive) approach for P-CBS that neglects the optimization with respect to prediction uncertainty is also discussed in this work. The P-CBS scheduler proposed in this paper is basically optimized for LoS fading and it achieves a better trade-off between sum-rate and fairness even in No-LoS when compared to non-predictive schedulers.