On Statistical Interference Modeling and Deployment Issues for Cognitive Radio Systems in Shadow Fading Environments

In this paper we deal with the issues of statistical modeling of the aggregate interference due to cognitive radios (CRs) at primary users and the deployment of CR systems in real world environments. We explore the possibilities of modeling the interference by two and three-parameter log-normal distributions. We find that the traditional approach of using a two-parameter log-normal is not a suitable option on account of the massive skewness difference between the actual interference and the log-normal variable used for fitting purposes. Similarly, we observe that the recently proposed three-parameter log-normal approximation works only under specific scenarios. While probing these issues, we also arrive at a considerably simpler way of deriving the cumulants of the total interference. Due to analytical difficulties, we resort to simulations to assess different CR deployment schemes. Based on CRs equipped with apriori knowledge of the radio environment map (REM), we compare various CR system design parameters achieved by the proposed techniques. Furthermore, we describe the effect of imperfections in the REM information on the performance of CR systems. Via analysis and simulations we explore the relationship between the required precision of the REM and various system properties. Index Terms Cognitive radio networks, interference statistics, fading channel, radio environment map, primary exclusion zone. A subset of this work was presented at the IEEE ICC [1] and IEEE AusCTW [2]