A clustering approach for admission control and optimal beamforming in cognitive radio networks

We propose a computationally efficient approach to the problem of admission control and beamforming in a cognitive radio network scenario in which a secondary base station selects the largest set of users to be optimally served with required quality of service while minimizing the transmit power and respecting the interference thresholds imposed by a dense incumbent primary system. To reduce the complexity of the combinatorial user selection problem we propose an initial user clustering scheme in which the similarity between the long term spatial signatures of the users is assessed. Based on these clusters, the users to be simultaneously served, along with their optimal transmit powers and beamforming vectors, are selected according to their instantaneous spatial signatures using an iterative uplink-downlink algorithm. Numerical results show a significant reduction in complexity of our approach compared to previous techniques.