Downlink Rate Maximization in FDD Operated Massive MIMO System with Delayed CSI Feedback

This paper considers a massive Multiple Input Multiple Output (MIMO) communication system in Frequency Division Duplexing (FDD) mode, where downlink rate performance imposes a considerable overhead of channel state information (CSI) feedback; thus, impacting the uplink rate of users. Hence we propose a delayed CSI feedback mechanism with the introduction of a metric, Age of Channel State Information (AoCSIT), characterizing the resulting staleness in channel state information at transmitter (CSIT). Moreover, considering a non-ideal feedback channel and correlated downlink channels across frames, we express the achievable downlink rate as a function of AoCSIT and formulate an optimization problem to maximize the sum downlink rate with constrain on the user’s minimum required uplink data rate. Next, we decouple the optimization problem and reformulate it as a Constrained Markov Decision Process (CMDP) based optimization problem. We solve this by forming Lagrangian, followed by a primal-dual approach. The primal problem, an unconstrained Markov Decision Process (MDP), is solved by Relative Value Iteration Algorithm (RVIA). However, to manage the computational complexity of the RVIA, due to large state space, we exploit its structural properties and propose an optimal deterministic threshold policy. We demonstrate the overall efficacy of the proposed solution through numerical simulations.