Decentralized Precoding for Multicell MIMO Downlink

Interference is a performance limiting factor in dense cellular networks with aggressive frequency reuse. Cooperation among base stations (BSs) is a promising approach for improving data rates by eliminating or mitigating interference. For the downlink, the highest spectral efficiency gains are achieved through precoding with full coordination, which requires complete channel state information (CSI) and data be shared among BSs at the cost of significant utilization of the backhaul. In this paper, we propose distributed precoding techniques for the multicell MIMO downlink. Unlike prior work, our proposed precoders are both decentralized with respect to the BSs as well as capable of enabling multiple users to share the same frequency carrier spatially within each cell. Specifically, each BS designs its own precoder without requiring data or downlink CSI of links from other BSs. Since CSI is unlikely to be perfect, we study the effect of imperfect CSI on our proposed precoders and propose a robust precoder in the presence of CSI uncertainty. Simulations show that our proposed methods enjoy a rate increase with SNR similar to multicell joint dirty paper coding in the high SNR regime due to effective interference mitigation. Numerical results reflect the sensitivity of each proposed precoder with respect to the imperfectness in the available CSI.