Grassmannian Training for Massive MIMO Cellular Networks

Massive multiple-input multiple-output (MIMO) systems are one of key technologies for next generation cellular providing high spectral efficiency. While the effect of most interference and noise vanishes as the number of antennas increases, performance of massive MIMO systems is limited by pilot contamination caused by correlated pilot. Pilot reuse, allowing users in distant cells to use the same pilot, has been recently proposed to mitigate the pilot contamination with a reasonable training overhead. However, existing pilot reuse schemes impose orthogonality constraint on the different pilot sequences, resulting in inflexibility in optimizing the training period. In this paper, we investigate the use of pilot of an arbitrary length in the pilot reuse framework. We also show that the Grassmannian subspace packing can be leveraged to find an optimal training sequence for regular hexagonal cellular networks adopting pilot reuse.