[Paper Review] Constructive Interference in Linear Precoding Systems: Power Allocation and User Selection

Dimitris Christopoulos, Symeon Chatzinotas+2 more|arXiv (Cornell University)|Mar 29, 2013

Advanced MIMO Systems Optimization1 references3 citations

TL;DR

This paper proposes optimized power allocation and a low-complexity user selection algorithm for constructive interference zero forcing (CIZF) precoding in MIMO broadcast channels. By selectively preserving beneficial multiuser interference and using a semi-parallel user selection (SPUS) heuristic, the system achieves up to 2 dB SNR gain over conventional methods, with performance within 1 dB of optimal selection and minimal fairness degradation.

ABSTRACT

The exploitation of interference in a constructive manner has recently been proposed for the downlink of multiuser, multi-antenna transmitters. This novel linear precoding technique, herein referred to as constructive interference zero forcing (CIZF) precoding, has exhibited substantial gains over conventional approaches; the concept is to cancel, on a symbol-by-symbol basis, only the interfering users that do not add to the intended signal power. In this paper, the power allocation problem towards maximizing the performance of a CIZF system with respect to some metric (throughput or fairness) is investigated. What is more, it is shown that the performance of the novel precoding scheme can be further boosted by choosing some of the constructive multiuser interference terms in the precoder design. Finally, motivated by the significant effect of user selection on conventional, zero forcing (ZF) precoding, the problem of user selection for the novel precoding method is tackled. A new iterative, low complexity algorithm for user selection in CIZF is developed. Simulation results are provided to display the gains of the algorithm compared to known user selection approaches.

Motivation & Objective

• To address the lack of power allocation optimization in constructive interference zero forcing (CIZF) precoding systems, which previously assumed equal power allocation.
• To improve spectral efficiency and fairness in CIZF by jointly optimizing power allocation and user selection.
• To develop a low-complexity, iterative user selection algorithm that approaches optimal performance without exhaustive search.
• To investigate the impact of user selection on CIZF performance, especially in relation to constructive interference exploitation.
• To evaluate the trade-off between sum-rate maximization and fairness in CIZF systems under optimized power and user selection.

Proposed method

• Proposes a power allocation strategy that maximizes either sum-rate (throughput) or minimum user rate (fairness), using convex optimization techniques.
• Introduces a novel semi-parallel user selection (SPUS) algorithm that iteratively selects users based on channel gain and real-valued cross-correlation matrix elements.
• Uses the matrix $\mathbf{G} = \text{diag}(\mathbf{s}) \cdot \mathfrak{Re}(R) \cdot \text{diag}(\mathbf{s})$ to model constructive interference effects and guide user selection.
• Applies a greedy selection rule: at each step, the user with the highest contribution to the selected user’s effective channel gain is chosen, while maintaining orthogonality constraints.
• Implements a pruning mechanism to exclude already-selected users and iteratively builds the user set $\mathcal{S}_{(i)}$.
• Applies a final thresholding step to set negative cross-correlation terms in the output precoder to zero, ensuring only constructive interference is preserved.

Experimental results

Research questions

• RQ1 What is the impact of optimized power allocation on the sum-rate and fairness performance in CIZF precoding systems?
• RQ2 How does user selection affect the performance of CIZF precoding, and can a low-complexity algorithm achieve near-optimal results?
• RQ3 Can constructive interference be leveraged in a practical, non-linear precoding framework without requiring full channel state information or complex signaling?
• RQ4 What is the trade-off between sum-rate maximization and fairness when combining power allocation and user selection in CIZF?
• RQ5 How does the size of the user pool affect the performance gain from user selection in CIZF systems?

Key findings

• Optimizing power allocation in CIZF precoding yields a gain of over 2 dB in SNR compared to equal power allocation, significantly improving sum-rate performance.
• The proposed SPUS algorithm achieves performance within less than 1 dB of the optimal user selection, even with a finite user pool of 12 users.
• When combined with max-min fairness power allocation, the SPUS algorithm maintains low variance in user rates, indicating minimal compromise on fairness.
• System performance saturates for small user pools (e.g., K=12), indicating that most gains are achieved with practical-sized user sets.
• The simulation results confirm that constructive interference can be effectively exploited in linear precoding, and that user selection plays a critical role in maximizing its benefits.
• The SPUS algorithm reduces complexity compared to exhaustive search while maintaining close-to-optimal performance, making it suitable for real-time implementation.