Optimal user scheduling and power control in multi-user cognitive broadcast systems

Cognitive radio systems should not only have the ability to sense and exploit "frequency spectrum holes", but also the ability to sense and utilize "spatial spectrum holes". In this paper, we consider the issue of maximizing the throughput of the cognitive systems by fully utilizing "spatial spectrum holes" brought in by multiple cognitive users, in the scenario where a pair of licensed users and a cognitive broadcast system share multiple spectrum bands. By exploiting the channel reciprocity under the premise that the licensed system adopts the time-division-duplexing (TDD) mode, we propose a more practical cognitive access scheme that can sense the interference at the licensed user caused by the cognitive transmitter, based on the existing feedback signals from the licensed user to the licensed base station. Taking both interferences from the licensed base station to the cognitive receiver and from the cognitive transmitter to the licensed user into consideration, we investigate the optimal user scheduling and power allocation scheme that can maximize the ergodic sum rate of the cognitive system. We show that scheduling the user whose channel gain to interference and noise ratio (CGINR) is the largest for each frequency band is optimal. We also derive the dynamic power allocation scheme meeting the three practical constraints, i.e., the transmitter's average transmission power constraint, the power amplifier's instantaneous transmission power constraint, and the interference power constraint at the licensed user. The result shows that in different coherent time intervals and different frequency bands, the power allocation has a multi-level waterfilling structure. Theoretical analysis shows that the strategy scheduling user with the largest CGINR introduces significant performance improvement compared with the traditional strategy scheduling user with the largest channel gain to noise ratio (CGNR). We also illustrate the impact of power constraints and the number of users on system performance by simulation.