Online Power Allocation for Sum Rate Maximization in TDD Massive MIMO Systems

In this paper, we investigate the power allocation problem with perfect channel state information (CSI) for down-link sum rate maximization in time duplex division massive MIMO systems. We note that the downlink sum rate is generally a non-convex function of the power allocation vector and the corresponding solution space increases exponentially with the number of simultaneous users, which make the optimal power allocation computationally intractable in general. Here, we introduce an online paradigm to achieve a tradeoff between computational complexity and sum rate performance, which utilizes the state-of-art online learning techniques to iteratively update the power allocation according to continuous CSIs. The computational complexity of the proposed algorithm is highly reduced by using the first order optimization technique and the system sum rate is improved by exploiting the time correlation of wireless channels. Simulation results show that the downlink sum rate can be increased by 15%-100% by using our proposed algorithm, compared with the conventional average and pathloss-based power allocation schemes.