A channel estimation framework for high-mobility FDD massive MIMO using partial reciprocity

The estimation of Channel State Information (CSI) is one of the most difficult tasks for massive multiple-input multiple-output (MIMO) in frequency division duplex (FDD) mode. It is even more challenging in high-mobility scenarios. In this paper, we consider an FDD massive MIMO system with high-mobility and CSI delay and aim to predict the downlink (DL) channel under a realistic multipath channel model. The key novelty lies in the fact that for the first time we devise a joint angle-delay-Doppler (JADD) channel estimation framework. The main idea of our framework is to reconstruct the DL channel with the DL channel parameters estimated from the uplink (UL) channel samples and scalar feedback coefficients. To alleviate the feedback overhead, we design a wideband beamformer for the base station (BS) based on the DL angle-delay-Doppler parameters. The user equipment (UE) then estimates the DL channel parameters and feeds back some Doppler-related scalar coefficients back to the BS. We show that the feedback and DL pilot training overhead are independent of the number of BS antennas. The lower bound performance of our framework is also derived. Numerical results under the industrial channel model in rich scattering environments demonstrate that our framework works well from medium mobility scenario of 30 km/h to high mobility settings of 350 km/h.