Distributed Gradient Based Gain Allocation for Coherent Multiuser AF Relaying Networks

A set of distributed non-cooperating relay antennas is known to be capable of orthogonalizing multiple source- destination pairs in space by coherent amplify-and-forward (AF) relaying techniques. Known relay gain allocation schemes for this setup require either global knowledge of the channel coefficients between all participating nodes or an excess number of relays. The required dissemination of the channel state information (CSI) introduces a major overhead, which in practice may well diminish the spatial multiplexing gain. We introduce a new distributed gradient based gain allocation scheme, which minimizes this overhead. Key to this is the proof, that the gradient of the destination signal-to-interference-plus-noise ratio (SINR) can be calculated in a distributed manner based on local CSI at the relays and very limited feedback from the destinations. In order to minimize the number of iterations in the distributed gradient algorithm we propose two distributed approaches to determine the optimal step size for each iteration. Finally we provide simulation results on the sum rate performance and identify the sequential opening of the spatial channels as a key factor that impacts speed of convergence.