Abstract
In this paper we have considered optimal power allocation for a system model of multiple secondary user (SU) pairs coexisting with a single primary user (PU) pair in a spectrum underlay cognitive radio network. A new Genetic algorithm (GA) based power allocation scheme has been proposed to achieve maximum system utility subject to PU interference constraint and satisfaction of QoS requirement of all the SUs. The GA method is capable of power allocation with much less computational time. The scheme is further tested with a modified system utility function that uses reduced number of constraints, by minimizing the aggregate primary interference power. This approach, besides being faster, has the advantage of significantly reducing the interference power, thereby allowing more number of secondary users. The performance comparisons of the two utility functions have been reported with respect to relevant system parameters. The accuracy and convergence performance of our proposed strategy matches very closely to that of geometric programming.