Spectrum sharing in cognitive radio with quantized channel information

In this paper, we consider a wideband spectrum sharing system where a secondary user can share orthogonal frequency bands with a number of primary users, each licensed to use an individual band. We address the problem of ergodic capacity maximization of the secondary user subject to an average sum (across the bands) transmit power constraint on the secondary user and individual average interference constraints on the primary users. After studying optimal power allocation with full channel state information (CSI) at the secondary transmitter, we focus on designing optimal power allocation strategies for the quantized CSI case. It is assumed that a band manager has the full CSI of all secondary-to-secondary and secondary-to-primary channels, and designs (ofine) an optimal power codebook based on statistical information of the channels and feeds back the index of the codebook to the secondary transmitter for every channel realization. A modified generalized Lloyd type algorithm (GLA) is designed for an optimal power codebook. With the narrowband case (one secondary and one primary user) in mind, various properties of the quantized powers are derived. It is observed that only 2-3 bits of feedback with this codebook is sufficient to provide secondary capacities close to the full CSI case.