Improved channel estimation for massive MIMO systems using hybrid pilots with pilot anchoring

Abstract

Pilot contamination is an impairment in massive multiple-input multiple-output (MIMO) systems that introduces interference in both the uplink and downlink. Existing schemes for channel estimation employ time-multiplexed pilots, which require dedicated symbols for pilot training, and therefore have to be reused across cells. Superimposed pilots, on the other hand, require no overhead and offer a larger set of pilots that can be reused over larger number of cells, thereby offering better performance in scenarios with high inter-cell interference. However, in scenarios with low inter-cell interference, the data that is transmitted alongside the pilots causes self-interference, which limits the performance. In this talk, we consider a multi-user massive MIMO system wherein these time-multiplexed pilots are augmented with superimposed pilots in the uplink data transmission phase. For this system, we obtain the channel estimate by imposing a shape constraint on the least-squares (LS) estimator for superimposed pilots. The LS estimator is constrained such that the resulting channel estimate, when used in a matched-filter, retrieves the known transmitted data (time-multiplexed pilots, in this case) from the received observations exactly. This optimization problem is then shown to be analogous to the generalized sidelobe canceller and a closed form expression for the channel estimate is obtained. Based on simulation results, the proposed channel estimation method is shown to improve the performance in high-interference scenarios when compared to channel estimators that employ only time-multiplexed pilots or only superimposed pilots. In addition, in low-interference scenarios, the shape constraint results in a reduction in the self-interference, thereby improving performance.