Distributed Sum Rate Maximization Algorithms for Nonregenerative MIMO Relay Networks

Distributed algorithms are critical for practical applications. Compared to centralized algorithms, they save much computation time and signaling overhead. We consider a two-hop network with multiantenna user pairs and multiantenna relays and propose two distributed algorithms to update the precoding, the decoding, and the relay amplify-and-forward matrices based on local channel state information. Both algorithms assume conferencing relays. One algorithm maximizes an approximation of the sum rate, where the convergence of the objective function is guaranteed. The other algorithm considers user pairs with limited computational capabilities, and relays take most calculations. The overhead and the distributed implementations are analyzed. Simulation results show that the two algorithms improve the existing distributed algorithm significantly in terms of the sum rate; both of them achieve at least one half of the sum rate by the centralized algorithm, but with much less complexity and lower overhead. The two algorithms provide diverse operating points for different types of networks, by well balancing the tradeoff among the achieved sum rate, the computational cost, and the feedback overhead.

I. Introduction

Many works about multiple-input multiple-output (MIMO) networks assisted by amplify-and-forward (AF) relays have studied the design of precoding, decoding, and relay AF strategies [1]–[23]. Recent work [20] is based on general MIMO AF relay networks with multiple links and multiple relays. Weighted minimum mean-square-error (WMMSE) model is proposed. The authors of [21]–[23] propose the total signal to total interference plus noise ratio (TSTINR) approach to approximate the sum rate maximization problem.