Decentralized spectrum learning and access adaptive to channel availability distribution in primary network

Abstract

We consider the effect of the mean availability distribution of primary channels on the performance of distributed learning and access policies, and develop a distributed learning and access policy that is effective in a wide range of primary channel conditions. We first extend the recently proposed BLA algorithm to distributed online learning of underlying primary channel availabilities, and modify the existing access policies to form BLA-ρ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RAND</sup> and BLA-DLF policies. By analyzing the distributed access collision mechanism offered by the ρ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RAND</sup> and DLF policies [1], [2], we identify how different mean channel availability distributions can impact the effectiveness of each policy. In light of this, we propose DSLA policy that adapts to different channel availability distribution conditions. Based on a closeness factor we propose, the DSLA policy automatically switches between the underlying learning policies, as well as the access policies, to determine which policy is most effective for a given primary channel condition. Simulation studies show that our proposed DSLA policy is effective in providing a good performance for a wide range of primary channel availability distributions.