Reconfigurable Intelligent Surfaces-Enabled Intra-Cell Pilot Reuse in Massive MIMO Systems

TL;DR

This work tackles pilot contamination in dense massive MIMO by enabling intra-cell pilot reuse through RISs. It introduces a statistical-CSI–driven RIS phase-shift optimization and a deterministic RIS placement strategy to group UEs served by different RISs, reducing interference while preserving channel gain. A MMSE-based channel estimator and standard MR/ZF/MMSE combiners/precoders are used to assess performance, with extensive UL and DL simulations showing substantial spectral efficiency gains—up to around 64% in UL and significant improvements in DL—compared with no RIS or randomly phased RIS configurations. The results demonstrate that RIS-aided intra-cell pilot reuse can dramatically cut training overhead and mitigate pilot contamination, offering a practical path toward scalable 6G mMIMO deployments. Possible future directions include UE-RIS clustering and near-field RIS behavior.

Abstract

Channel state information (CSI) estimation is a critical issue in the design of modern massive multiple-input multiple-output (mMIMO) networks. With the increasing number of users, assigning orthogonal pilots to everyone incurs a large overhead that strongly penalizes the system's spectral efficiency (SE). It becomes thus necessary to reuse pilots, giving rise to pilot contamination, a vital performance bottleneck of mMIMO networks. Reusing pilots among the users of the same cell is a desirable operation condition from the perspective of reducing training overheads; however, the intra-cell pilot contamination might worsen due to the users' proximity. Reconfigurable intelligent surfaces (RISs), capable of smartly controlling the wireless channel, can be leveraged for intra-cell pilot reuse. In this paper, our main contribution is a RIS-aided approach for intra-cell pilot reuse and the corresponding channel estimation method. Relying upon the knowledge of only statistical CSI, we optimize the RIS phase shifts based on a manifold optimization framework and the RIS positioning based on a deterministic approach. The extensive numerical results highlight the remarkable performance improvements the proposed scheme achieves (for both uplink and downlink transmissions) compared to other alternatives.

Figures (11)

• Figure 1: Multiuser mMIMO communication system assisted by multiple RISs, each deployed on the facades of buildings. The users in the dotted circle area are served without the aid of any RIS. They, as well as the users served by each RIS, share the same set of pilot sequences $\{ \psi_1, \psi_2, \psi_3\}$ in our investigated scenario. Thus, each user sees interference of other users reusing the same pilot (IPR), as well as interference of users using other pilots (IOP).
• Figure 2: Optimized grid of angular positions for the RIS deployment: a) $M=16$; and b) $M=128$ BS antennas. The BS boresight is the horizontal axis.
• Figure 3: Average normalized interference level between channel vectors of UEs aided by different RISs. We denote by "opt" when the RIS deployments positions are fitted to the optimized angular grid proposed here. Besides, the grey scale for the "mo" curves scales according to the Rician factor $\alpha_r^{\textsc{br}} \in [0, 3, 5, 10]$ dB. The blue rectangle in a) is expanded in b), where the green arrows signalize the trend of the curves with the increasing Rician factor.
• Figure 4: Spatial topology of the network when $\varsigma = 4$, and $R=3$: a) $K=4$, $\tau_p = 1$; and b) $K=16$, $\tau_p = 4$.
• Figure 5: UL SE with increasing $M$, when $R=3$, $N=256$, and $K=4$ UEs share the same pilot.