Mean-Variance Based QoS Management in Cognitive Radio

Abstract

Guaranteeing Quality of Service (QoS) in Cognitive Radio (CR) networks is a challenging task due to the random nature of radio channel conditions and primary user traffic. In this paper we analyze the recently proposed mean-variance based QoS and resource management methods, and introduce the concept of mean-variance evaluation of QoS and resource management techniques. Inspired by financial Portfolio Selection theory, mean-variance based resource management techniques for CR networks are statistical approaches and do not require instantaneous knowledge of channel state and Primary User (PU) activity, and enable the option to tradeoff between risk (QoS variance) and reward (QoS mean). Using throughput as a measure of QoS, we present a derivation of the theoretical throughput mean-variance characteristics of a CR-OFDM system employing a mean-variance based QoS management strategy. We conduct an analysis of existing Portfolio Selection based strategies in the mean-variance domain and compare them to the theoretical performance. Finally, we present a further enhancement of the approach by explicitly considering constraints on individual channel power allocation in the mean-variance optimization problem. Simulation results illustrate the effect of our enhancements in improving the risk-reward profile of the mean-variance based QoS management strategy, by moving it closer to the theoretical characteristic.