Distributed LMMSE estimation in cooperative cellular networks

We consider a cooperative cellular network with multiple users transmitting within a cooperative cluster of multiple antenna base stations. Different from global cooperation where all the signal processing is performed at a central processor, we analyze a distributed processing architecture that performs the estimation of user symbols in two steps, namely, 1) Local LMMSE estimation of the user symbols at each base station, and 2) Central combining of all the estimates from the base stations. For such a network, we derive new expressions for the capacity of a given user with independent Rayleigh fading and arbitrary path loss between the users and all the antennas at the base stations. Our capacity expression is based on accurate approximations we derive for the characteristic function and the probability density function of the users´ signal-to-interference-plus-noise ratios (SINRs). Numerical examples demonstrate that our analytical solutions accurately approximate the exact capacity. Furthermore, we highlight that the distributed approach introduces less overhead to the network compared with global cooperation. We note that the performance gap is small when the network is lightly loaded.