Comparative analysis of GA and SA for utility maximization of licensed and unlicensed users in a cognitive radio network

In this paper we consider a cognitive radio network comprising of a number of primary users (PUs) and a number of secondary users (SUs). The spectrum has been divided into channels by means of frequency division multiple access (FDMA). The PUs have license to use the channels and the SUs periodically check the channels for idleness. When the channels are not in use by the PUs, the SUs may bid for them. The PUs auctions off the channels to the purchasers (SUs) who offer most money for them. Thus, PUs earn revenue in place of the spectrum. Here, we intend to compute an allocation of channels/vacant spectrums to the SUs such that the total revenue earned by all PUs is maximized. We also intend to compute another such allocation of channels/vacant spectrum to the SUs such that the payoff earned by the SUs is maximized. Both these optimizations have been performed with the help of real coded genetic algorithm (GA) and simulated annealing (SA) algorithm. We have made comparative analysis of both the algorithms and show that GA provides better near optimal solution compared to SA in terms of computation of a close to optimal solution. But, computational time of GA is more than that of SA. GA suffers from premature convergence, which can be dealt with SA.