Decontaminating pilots in cognitive massive MIMO networks

Abstract

Cognitive radio has been lately suggested as a promising technology in order to improve spectrum utilization. This paper addresses the problem of channel estimation in an underlay interference-prone cognitive radio setup. We consider a primary and a secondary base station, both with multiple antenna capability and serving multiple users. Although previous studies propose the use of beamforming to handle secondary-caused interference, this cannot be done in practice unless channels are correctly estimated in the first place. However channel estimation itself is plagued by interference (pilot contamination effects). Therefore we propose a method to address channel estimation at the primary system while removing contamination caused by the secondary transmitter. The approach is twofold: (i) We develop a robust channel estimator which makes use of covariance information. We show analytically that the performance of this estimator is identical to the interference free scenario under certain when the number of antennas becomes large under a condition on the distribution of the multipath model. (ii) We build a pilot assignment algorithm which seeks to fulfill this condition. Significant gains are reported.