Performance of Multi-Cell Massive MIMO Systems With Interference Decoding

We consider a multi-cell massive MIMO system where a time-division duplex protocol is used to estimate the channel state information via uplink pilots. When maximum ratio combining (MRC) is used at the BSs, the re-use of pilots across cells causes the pilot contamination effect which yields interference components that do not vanish as the number of base-station (BS) antennas <inline-formula> <tex-math notation="LaTeX">$M \rightarrow \infty $ </tex-math></inline-formula>. When treating interference as noise (TIN), this phenomenon limits the performance of multi-cell massive MIMO systems. In this paper, we analyze more advanced schemes based on simultaneous unique decoding (SD) as well as simultaneous non-unique decoding (SND) of the interference that can provide unbounded rate as <inline-formula> <tex-math notation="LaTeX">$M \rightarrow \infty $ </tex-math></inline-formula>. We also establish a worst-case uncorrelated noise technique for multiple-access channels to derive achievable rate expressions for finite <inline-formula> <tex-math notation="LaTeX">$M$ </tex-math></inline-formula>. Furthermore, we study a much simpler subset of SND (called S-SND) which provides a lower bound to SND and achieves unbounded rate as <inline-formula> <tex-math notation="LaTeX">$M \rightarrow \infty $ </tex-math></inline-formula>, and also outperforms SD for finite <inline-formula> <tex-math notation="LaTeX">$M$ </tex-math></inline-formula>. For the special cases of two-cell and three-cell systems, using a maximum symmetric rate allocation policy we compare the performance of different interference decoding schemes with that of TIN. Finally, we numerically illustrate the improved performance of the proposed schemes.