Robust worst-case design for optimizing average performance in OFDM using quantized CSI

Orthogonal frequency-division multiplexing (OFDM) has been able to successfully exploit channel state information (CSI) at the transmitter, allowing to implement dynamic resource allocation schemes that improve spectral efficiency and error resilience. In most wireless communication systems, though, achieving a perfect CSI at the transmitter is difficult. For this reason a limited-rate feedback mode has been proposed, in which only quantized CSI at the transmitter is available through a finite number of bits that are fed back by the receiver. In this paper we design resource allocation schemes for OFDM systems, that use limited-rate feedback and do not assume any structure on the channel quantizer. The new schemes are obtained by solving an optimization problem that maximizes average ergodic rate subject to average power and bit error rate constraints. The latter constraint is satisfied by implementing a worst-case robust approach that reduces the dimensionality of the problem. The main burden is associated to the computation of optimal Lagrange multipliers. Provably convergent ensemble schemes that rely on the channel distribution as well as stochastic schemes that catch the average behavior of the system on-the-fly are developed for such task. Different alternatives to reduce the amount of feedback and the extension to multiple user systems wrap-up this paper.