Enhancing massive MIMO: A new approach for Uplink training based on heterogeneous coherence time

Massive MIMO is one of the key technologies in future generation networks. Owing to their considerable spectral and energy efficiency gain, massive MIMO systems provide the needed performance to cope with the ever increasing wireless capacity demand. Nevertheless, the number of scheduled users stays limited in massive MIMO systems both in TDD and FDD modes. This is due to the limited time slot duration in TDD mode and to the limited feedback capacity in FDD mode. In current systems, the time slot duration in TDD mode is the same for all users. This is a suboptimal approach since users are subject to heterogeneous Doppler frequencies and, consequently, different coherence times. In this paper, we argue that, exploiting the diversity in the users coherence time in order to optimize the training procedure can lead to considerable spectral efficiency gain. We investigate a massive MIMO system operating in TDD mode in which, the frequency of uplink training differs among users based on their actual channel coherence time. We then provide a user scheduling algorithm that exploits a coherence interval based grouping in order to maximize the achievable weighted sum rate.