Abstract
Recent works on multi-user transmission techniques have shown that the linear growth of capacity in single user MIMO systems can be translated to the multi-user MIMO scenario as well. In this paper, we propose a method pursuing performance gain of vector perturbation in multi-user downlink channels. Instead of exploiting the maximum available mobile users for communication, we employ a small part of them as "sacrifice lambs" for improving the quality of service (QoS) of active users. By controlling the information and perturbation vectors of these sacrificing users, we achieve considerable improvement in the effective SNR, resulting in a large gain in bit error performance. Simulation results show that the proposed method brings substantial performance gain over standard vector perturbation.
Introduction
Multi-user Multiple-Input Multiple-Output (MU-MIMO) is a cornerstone of modern wireless standards, allowing a base station to communicate with multiple users simultaneously. To manage inter-user interference, Vector Perturbation (VP) is often employed. VP is a powerful non-linear precoding technique that adds a perturbation to the signal to make it easier for the channel to handle, effectively increasing system capacity.
However, VP has a major reliability issue: the performance of all users is tied to a single scaling factor ($\gamma$). If even one user has a very poor, "ill-conditioned" channel due to fading, the scaling factor spikes, effectively destroying the Signal-to-Noise Ratio (SNR) for every user on the network. This paper addresses this "weakest link" problem by introducing a strategy that selectively sacrifices users with poor channels to boost overall system reliability.
Proposed Methodology: The Virtual User Strategy
The authors introduce a user selection and precoding framework based on the concept of Virtual Users. The process follows these key steps:
- Outage Detection: The system identifies users in an "outage" state—those whose potential data rate falls below a predefined threshold ($\epsilon$).
- Virtual User Conversion: Instead of simply dropping these users, the system treats them as "virtual users." Their channel vectors remain part of the mathematical model, but instead of carrying their own data, their signal components are treated as free variables used to optimize the system.
- Hybrid Optimization: Standard VP requires finding an integer perturbation vector. By adding virtual users, the optimization becomes a hybrid problem: finding integer perturbations for active users and continuous (non-integer) values for virtual users to minimize the scaling factor $\gamma$.
- Modified Sphere Encoding: To solve this efficiently, the authors propose a modified sphere encoder. By projecting the signal onto the null space of the virtual users' channels, the search space is reduced, which can actually lower computational complexity compared to standard VP.
Performance Evaluation
The proposed method was evaluated using numerical simulations of a $4 \times 4$ MU-MIMO downlink system under Rayleigh flat fading conditions.
As illustrated in the BER results, the proposed method provides a significant performance boost. At a Bit Error Rate (BER) of $10^{-2}$, the method achieves a 2.5 dB gain over standard Vector Perturbation. This confirms that the "sacrifice for the greater good" strategy effectively stabilizes the scaling factor and improves reliability for active users.
There is a trade-off involved: because fewer users are receiving data, the total sum rate is slightly reduced. At an SNR of 20dB, the sum rate loss was approximately 12%. However, this gap narrows as SNR increases, suggesting that the massive jump in reliability is a fair price for a manageable hit to total throughput.
Conclusion
This paper demonstrates that by intentionally sacrificing users in deep fades and treating them as virtual optimization variables, MU-MIMO systems can achieve much higher reliability. While this introduces a minor reduction in total sum rate, the gain in BER performance and the potential for reduced computational complexity make it a robust choice for practical system design. Future work will need to address user fairness to ensure sacrificed users are eventually served through intelligent scheduling.
Summary updated