CDF Scheduling Methods for Finite Rate Multiuser Systems With Limited Feedback

摘要

In this work, simple and practical CDF scheduling methods are developed that preserve the virtues of CDF scheduling, namely fairness and effective use of multiuser diversity. They are the accelerated extended CDF scheduling (AeCS), modified non-parametric CDF scheduling (modified NPCS), and the CDF scheduling with optimized Quantizers (CSwQ) methods. The developed methods do not need knowledge of the channel distribution and can work effectively in systems that support discrete number of transmission rates and are constrained by limited feedback resources. They exploit the limited feedback resource for their scheduling decisions and transmission rate selections and learn the distribution of channel quality information of each user, as needed, to best exploit multiuser diversity. The CSwQ method is shown to have desirable properties in terms of both system throughput and power savings. Analytical results for the system throughput and power savings of the CSwQ method are developed that are valid for a finite number of observation samples. Then, through numerical simulations, the developed methods are shown to effectively save transmit power and achieve high system throughput. In addition, they converge rapidly with the number of data samples to the ideal case wherein perfect knowledge of the channel CDF is assumed.