Lattice coding for the vector fading paper problem

Dirty paper coding (DPC) is a promising preceding technique for canceling arbitrary interference known only at the transmitter. The interference-free rate is thus achieved. However, this result relies on the perfectly known channel coefficients at the transmitter when being applied to the fading channel. We thus consider the fading paper problem where only the channel statistics information is available at the transmitter. In general, the optimal transmission strategies to achieve the capacity of this channel is still unknown. We confine ourselves to the linear-assignment Gel'fand-Pinsker coding which has been proved to have good, sometimes even optimal, performance in the a variety of fast and slow fading channels. However, lack of structured codebook so far limited the practical applications of this coding. In this paper, we present a lattice-based coding structure for the vector fading paper channel. It can achieve the rate performance of the linear-assignment strategies previously proved by the random Gaussian codebook. Moreover, the lattice codebook has an algebraic structure and is possible to be implemented in practice. The results can be applied to the emerging fields such as fading multiple-input multiple-output (MIMO) Gaussian broadcast channels or fading MIMO cognitive channels.