Pilot-assisted ergodic interference alignment for wireless networks

This paper considers the ergodic block fading multi-user Gaussian interference channel (IC) in which each source desires to communicate to an intended destination. We assume that there is no CSI a priori available at terminals. We develop achievable rate results and compute the associated degrees of freedom by using a pilot-assisted interference alignment scheme. In this scheme, each source first sends known pilot symbols via which the destinations estimate channel gains, and the destinations then broadcast the estimated channel gains via orthogonal feedback channels. The estimated channel gains are used to perform interference alignment for data transmission. The pilot transmission power can be different from the data transmission power. By allocating more power to pilot transmission, channel gains can be estimated more accurately which implies less power left for data transmission. We find the optimum power allocation to pilot symbols and data symbols. Our study recommends, in large networks, to allocate more power to channel training instead of data transmission. In addition, our results reveal that for a K-user ergodic IC with a coherence time T, the total degrees of freedom 1/2 K_opt (1-K_opt/T) is achievable, where K_opt = min {K, T/2} is the optimum number of users selected to be active in the network. This recommends to perform a user selection in large networks (K > T/2), and apply channel training and interference alignment within the set of selected users.