IEEE Transactions on Wireless Communications, vol. 10, no. 11, pp. 3570-3576, November 2011.
We propose transceiver algorithms in cognitive radio networks where the cognitive users are equipped with multiple antennas. Prior work has focused on a design of precoding matrices to suppress interference to primary receivers. This work considers designs of precoding and decoding matrices for spatial sensing to achieve two objectives: i) to prevent interference to the primary receivers and ii) to remove the interference, due to primary transmissions, at secondary receiver. Three different scenarios, depending upon the amount of channel state information (CSI) at the secondary transmitter and receiver, are studied: i) local CSI, ii) global CSI, and iii) local CSI with side information. When the local CSI is available, for a simple design, we use the prior work, projected-channel singular value decomposition (P-SVD) technique. In the global CSI scenario, we propose a joint transmit-receive design under the assumption of full CSI of all users at the secondary transceiver. To overcome a problem of feedback overhead in the global CSI scenario, we propose a new iterative algorithm when the local CSI and side information are available. In this algorithm, the secondary transmitter and receiver update precoding and decoding matrices, respectively, to maximize the rate of the secondary link while maintaining the zero-interference constraint. When beamforming is used, convergence of the iterative algorithm is established. Numerical results show that the proposed joint design and the iterative algorithm show better achievable rate performance than the P-SVD technique at the expense, respectively, of CSI knowledge and side information.