Splitting Messages in the Dark- Rate-Splitting Multiple Access for FDD Massive MIMO Without CSI Feedback

A critical hindrance to realize frequency division duplex (FDD) massive multi-input multi-output (MIMO) systems is overhead associated with downlink channel state information at the transmitter (CSIT) acquisition. To address this challenge, we propose a novel framework that achieves robust performances while completely eliminating downlink CSIT training and feedback. Specifically, by exploiting partial frequency invariance of channel parameters between the uplink (UL) and downlink (DL), we adopt the 2D-Newtonized orthogonal matching pursuit (2D-NOMP) algorithm to reconstruct DL CSIT from UL training. Due to inherent discrepancies arising from a carrier frequency difference between two disjoint bands, however, the multi-user interference is inevitable. To overcome this, we propose a precoding method that employs rate-splitting multiple access (RSMA) and also develop an error covariance matrix (ECM) estimation method by using the observed Fisher information matrix (O-FIM). We find that this ECM estimation is crucial for our precoding design in maximizing the sum spectral efficiency (SE). Simulation results show that our method significantly improves the sum SE compared to other state-of-the-art approaches, underscoring the importance of our ECM estimation.