Performance Limits of a Cloud Radio

Cooperation in a cellular network is seen as a key technique in managing other cell interference to observe a gain in achievable rate. In this paper, we present the achievable rate regions for a cloud radio network using a sub-optimal zero forcing equalizer with dirty paper precoding. We show that when complete channel state information is available at the cloud, rates close to those achievable with total interference cancellation can be achieved. With mean capacity gains, of up to 200 % over the conventional cellular network in both uplink and downlink, this precoding scheme shows great promise for implementation in a cloud radio network. To simplify the analysis, we use a stochastic geometric framework based of Poisson point processes instead of the traditional grid based cellular network model. We also study the impact of limiting the channel state information and geographical clustering to limit the cloud size on the achievable rate. We have observed that using this zero forcing-dirty paper coding technique, the adverse effect of inter-cluster interference can be minimized thereby transforming an interference limited network into a noise limited network as experienced by an average user in the network for low operating signal-to-noise-ratios. However, for higher signal-to-noise-ratios, both the average achievable rate and cell-edge achievable rate saturate as observed in literature. As the implementation of dirty paper coding is practically not feasible, we present a practical design of a cloud radio network using cloud a minimum mean square equalizer for processing the uplink streams and use Tomlinson-Harashima precoder as a sub-optimal substitute for a dirty paper precoder in downlink.