Linear combining in dependent α-stable interference

Recently, there has been a proliferation of wirelesscommunication technologies in unlicensed bands for the Internetof Things. A key question is whether these networks can coexistgiven that they have different power levels, symbol periods,and access protocols. The main challenge is to characterizethe impact of mutual interference arising from distinct uncoordinatednetworks. It is known that when interferers forma homogeneous Poisson point process and transmit only on asingle subband, the interference is often well-modeled by theheavy-tailed α-stable distribution. In this paper, we focus onthe scenario where interferers transmit on multiple subbands.Under a policy where each interferer independently accesses eachband with probability p, we provide an exact characterization ofthe interference random vector. Exploiting this characterization,we derive optimal linear combining weights and an analyticalapproximation for the bit error rate (BER), accurate for largetransmit power. A key observation is that the expression for theBER admits an interpretation in terms of an array gain and afractional diversity gain.