Abstract
This paper considers the use of Non-Orthogonal Multiple Access (NOMA) in multiuser MIMO systems in practical scenarios where Channel State Information (CSI) is acquired through pilot signaling. A new NOMA scheme that uses shared pilots is proposed. Achievable rate analysis is carried out for different pilot signaling schemes including both uplink and downlink pilots. The achievable rate performance of the proposed NOMA scheme with shared pilot within each group is compared with the traditional orthogonal access scheme with orthogonal pilots. Our proposed scheme is a generalization of the orthogonal scheme and can be reduced to it when appropriate power allocation parameters are chosen. Numerical results show that when downlink CSI is available at the users, our proposed NOMA scheme outperforms orthogonal schemes. However, with more groups of users present in the cell, it is preferable to use multi-user beamforming instead of NOMA.
Introduction
In the evolution toward 5G and beyond, Multiuser MIMO (MU-MIMO) and Non-Orthogonal Multiple Access (NOMA) are two cornerstone technologies. While MU-MIMO leverages the spatial domain to separate users via large antenna arrays, NOMA utilizes the power domain through superposition coding and Successive Interference Cancellation (SIC).
Most theoretical evaluations of NOMA assume perfect Channel State Information (CSI). However, in practical massive MIMO deployments, channels must be estimated using finite pilot resources. This "pilot shortage" creates a bottleneck, especially as the number of users grows. This paper addresses whether NOMA maintains its performance edge when accounting for the overhead and estimation errors inherent in training-based systems.
System Model and Training Phases
The study focuses on a multiuser MIMO environment where the Base Station (BS) must estimate channels for multiple users. The process is divided into distinct phases: uplink training for BS estimation and downlink transmission for data delivery.
In the proposed NOMA scheme (Scheme-N), the system intentionally induces pilot contamination by assigning the same pilot sequence to a pair of users—typically one at the cell center and one at the cell edge. The BS uses Minimum Mean Square Error (MMSE) estimation to derive a combined channel estimate for the pair, as it cannot distinguish between them during the training phase. Subsequently, Maximum Ratio (MR) beamforming is applied to transmit superposed data to both users simultaneously.
Pilot Signaling and Frame Structure
To evaluate the impact of training overhead, the authors compare traditional orthogonal access (Scheme-O) with the proposed shared-pilot NOMA scheme. The frame structure is critical here, as it dictates how much of the coherence interval is dedicated to pilots versus data.
The analysis considers three specific CSI scenarios:
- Perfect CSI at the users: An idealized upper bound.
- No CSI at the users: Users rely on statistical channel averages for decoding.
- Estimated CSI at the users: Users estimate their effective beamformed channels via dedicated downlink pilots.
Key Insights and Results
The Criticality of Downlink CSI
A major finding is that NOMA's success is heavily dependent on the quality of CSI at the receiver. Without downlink pilots to help users estimate their effective channels, NOMA provides virtually no gain over orthogonal schemes. In such cases, the "self-interference" arising from estimation errors becomes the dominant bottleneck, negating the benefits of power-domain multiplexing.
The Near-Far Requirement
NOMA performs optimally when there is a significant disparity in path loss between the paired users (e.g., 20 dB). This "near-far" diversity allows the SIC process to effectively distinguish and cancel interference. When users have similar channel strengths, the spatial separation provided by MU-MIMO is generally more efficient.
Scaling Limits and Beamforming Dominance
While NOMA shows significant gains for a single pair of users—increasing the rate of a cell-center user by nearly 2 b/s/Hz in some scenarios—these gains diminish as more user groups are added. As inter-group interference begins to dominate, traditional multi-user beamforming with orthogonal pilots eventually outperforms the shared-pilot NOMA approach. This suggests that NOMA is most effective as a niche tool for specific "pilot-starved" scenarios rather than a universal replacement for spatial multiplexing.
Conclusion
This analysis demonstrates that while NOMA can mitigate pilot shortages through shared signaling, its practical utility is constrained by the need for high-quality downlink CSI and significant path-loss diversity. Future work may explore more advanced precoding techniques like Zero-Forcing (ZF) or the use of "blind" channel estimation methods to further reduce the overhead penalties associated with training-based NOMA systems.
Summary updated