Key Technologies Of Massive MIMO System

Massive MIMO (Multiple-Input Multiple-Output) system with a large number of an-tennas is considered in this thesis. Massive MIMO technology is motivated by the unprece-dented improvement of system performance, spectral and energy efficiency. Due to this new paradigm shift, Massive MIMO lends itself to a new class of challenges, which are com-pletely different from those familiar in the conventional small-scale MIMO. In recent years, the tremendous computational complexity of detection in Massive MIMO systems has been an active research area. In this thesis, we propose two near-optimal low-complexity detec-tion algorithms; the first is based on probabilistic graphical models, which is suitable for very large MIMO; the second is inspired by the stochastic algorithm named gravitational search algorithm, and it is suitable for a moderate number of antennas, a few tens.More recently, there has been a surge of interest on incorporating large antenna ar-rays in fifth generation cellular networks (5G), highlighting a new paradigm shift of theory, network design, and hardware implementation. One of the most difficult impediments in Massive MIMO is the acquisition of channel state information (CSI) at the base station (B-S), where the estimated CSI is, unfortunately, impaired by contamination from neighbouring cells, rendering such systems unable to realize their full promising gains. Recently, there has been an interest in blind channel estimation techniques, namely, subspace methods, to overcome pilot contamination problem in Massive MIMO.To that end, in this thesis, we leverage random matrix theory (RMT) and free prob-ability theory (FPT) to analyse the performance of such techniques under a more sensible physical channel model, which is more capable to capture the interaction between the an-tenna array and the physical environment, rather than the ideal assumption of independent and identically distributed (i.i.d.) Rayleigh channel adopted in the literature. Via asymptotic spectral analysis, it is demonstrated that the physical channel has, indeed, great impact on the performance of subspace methods. Furthermore, we demonstrate an interesting "anten-na saturation" effect, i.e., when the number of BS antennas approaches infinity, the perfor-mance saturates. In this thesis, we also propose a new channel estimation technique, which overcomes pilot contamination problem completely. The technique is based on superposi-tion of pilot and data-carrying signal, which is suitable for critical applications, where the degrees of freedom (DoF) of the channel are limited. In our algorithm, the high-quality performance and the required level of service are the main concerns.Finally, the non-coherent two-user interference channel (IFC) is investigated from the perspective of DoF. This research is, basically, a first step of understanding the non-coherent Massive MIMO channel, which is inherently a non-coherent IFC. New bounds to the sum DoF are derived when interference is assumed strong. The results show that the achievable DoF depends on the interference level in much the same way as in the coherent IFC, which is well-known in the literature. The obtainable results in this thesis imply that achieving interference-free DoF per user is possible from information-theoretic point of view, especially when interference is sufficiently large.