Interference Modeling and Avoidance in Spectrum Underlay Cognitive Wireless Networks

In spectrum underlay cognitive wireless networks, secondary nodes need to limit their aggregate interference on the primary receiving nodes. The trends for interference modeling has been either indiscriminate use of Central Limit Theorem to model the aggregate interference as a Gaussian random variable or by application of the Campbell's Theorem and approximating the probability density function of interference from its cumulants (e.g., by using Edgeworth or Gram-Charlier series). In the latter case, the theorem can be applied only when the interfering nodes have the same power level. In this paper, we deviate from the previous trends of interference modeling in following ways: (1) We allow the secondary neighbors of a primary node to have arbitrary power levels. (2) We split the set of interfering neighbors of a primary node into non-Gaussian (close neighbors) and Gaussian (far neighbors) interferers. For the case of Log-normal fading, we show that an accurate model for interference is sum of a Normal and a Log-normal random variables. We proceed to obtain an upper bound for the complementary cumulative distribution function of interference and show its tightness through simulation. Simulations results confirm the accuracy of the proposed model. Finally, we propose adjustable interference avoidance strategies and show that interference constraint is satisfied using these strategies.