Coordinated Pilot Design for Massive MIMO

Kaiming Shen, Yonina C. Eldar, Wei Yu

Introduction

In massive MIMO (Multiple-Input Multiple-Output) systems, base stations (BSs) require accurate Channel State Information (CSI) to perform beamforming and spatial multiplexing. This CSI is typically acquired through "pilots"—predefined sequences transmitted by users. However, the available coherence time limits the number of unique orthogonal pilots. In multi-cell networks, this leads to "pilot contamination," where users in different cells reuse the same pilots, causing interference that degrades channel estimates and creates a performance ceiling that more antennas cannot fix.

Conceptual comparison between (a) nonorthogonal and (b) orthogonal pilot schemes in a multi-cell network.

Traditional approaches attempt to mitigate this through greedy pilot assignment, but these often fail to reach global optima and struggle with the mathematical complexity of nonorthogonal sequences. This paper introduces FP-Pilot, a framework that moves beyond traditional orthogonal sequences to embrace coordinated, nonorthogonal pilot design.

Methodology: The FP-Pilot Framework

The authors treat pilot design as a Weighted Sum-Mean-Squared Error (MSE) minimization problem. Their core innovation is applying Matrix Fractional Programming (FP) to this non-convex challenge.

  1. Matrix Quadratic Transform: This mathematical technique decouples the numerator and denominator of the MSE expression, transforming a daunting non-convex problem into a sequence of simpler, convex sub-problems.
  2. FP-Pilot Algorithm: This iterative algorithm updates auxiliary variables and pilot symbols in closed-form. Unlike "greedy" methods that optimize one symbol at a time, FP-Pilot optimizes the entire pilot matrix jointly, ensuring faster convergence to a stationary point.
  3. Deterministic Rate Expression: The authors derived a new, closed-form expression for the achievable data rate with nonorthogonal pilots. This enables direct optimization without the need for complex expectations.
  4. Weight Optimization: Using the Lagrangian dual transform, the authors established a formal link between MSE minimization and rate maximization, providing a principled way to set weights to prioritize vulnerable cell-edge users.

Key Results and Performance

The framework was evaluated using numerical simulations of a 7-cell wrapped-around hexagonal network. Each cell featured a base station with 100 antennas serving 9 single-antenna users. The study compared FP-Pilot against standard orthogonal pilot reuse and state-of-the-art greedy algorithms.

Key findings include:

  • Massive Error Reduction: FP-Pilot reduced the sum-MSE by more than 9 dB compared to conventional orthogonal pilot-reuse approaches.
  • Throughput Gains: For users in the bottom 10th percentile (cell-edge users), the spectral efficiency was nearly 5 times higher than any existing method.
  • Rapid Convergence: Most performance gains (approx. 75%) were achieved within the first 10 iterations, making the algorithm computationally viable for practical deployment.
  • Hardware Compatibility: The coordinated design remained effective even when restricted to discrete constellations (like 4-QAM) for hardware simplicity.

Conclusion and Future Work

By embracing nonorthogonality, FP-Pilot provides a more flexible and robust approach to interference management in massive MIMO. However, the method requires coordination across multiple cells, implying a need for low-latency backhaul links. Future research will explore the overhead of re-calculating pilots in high-mobility scenarios and the performance of these designs under extreme hardware impairments like phase noise.

Summary updated

Abstract

Pilot contamination is a main limiting factor in multi-cell massive multiple-input multiple-output (MIMO) systems due to the non-orthogonality of pilot sequences which can seriously impair the channel measurement. Recent work has suggested coordinating pilot sequence design across multiple cells by choosing the sequences to minimize the channel estimation error. This paper further investigates this approach using a new optimization framework. Specifically, we reformulate the weighted minimum mean-squared error (MMSE) measure as a sum-of-functions-of-matrix-ratio program that can be efficiently solved via a matrix fractional programming approach. The proposed algorithm provides fast convergence to a stationary-point solution of the MMSE problem. Simulations demonstrate the advantage of the proposed method over alternative approaches in enhancing channel estimation accuracy.
BibTeX
@inproceedings{shen2019coordinated,
  author = {Kaiming Shen and Yonina C. Eldar and Wei Yu},
  title = {{Coordinated Pilot Design for Massive MIMO}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2019},
  doi = {10.1109/icassp.2019.8683449},
}

Loading related content...

Figure