Observation Matrix Design for Densifying MIMO Channel Estimation via 2D Ice Filling

2025-10-09

AI Summary

To address the limitation in channel estimation performance of massive MIMO systems caused by suboptimal observation matrix design, this paper proposes a joint transmitter-receiver observation matrix optimization framework leveraging prior knowledge of the channel covariance matrix. The key contributions are: (1) a two-dimensional ice-filling (2DIF) algorithm that decouples the channel covariance’s eigenspace into transmit and receive subspaces to construct near-optimal digital precoding and combining matrices; and (2) an extension to hybrid analog-digital architectures via a two-stage 2DIF (TS-2DIF) scheme, which jointly incorporates phase-shifter constraints and mutual information maximization. Simulation results demonstrate that the proposed methods significantly outperform state-of-the-art approaches in terms of normalized mean-squared error (NMSE) and estimation rate—particularly under low pilot overhead.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Search and Optimization: Mixed Discrete/Continuous Search
• Intelligent Robots: State Estimation

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

In recent years, densifying multiple-input multiple-output (MIMO) has attracted much attention from the communication community. Thanks to the subwavelength antenna spacing, the strong correlations among densifying antennas provide sufficient prior knowledge about channel state information (CSI). This inspires the careful design of observation matrices (e.g., transmit precoders and receive combiners), that exploits the CSI prior knowledge, to boost channel estimation performance. Aligned with this vision, this work proposes to jointly design the combiners and precoders by maximizing the mutual information between the received pilots and densifying MIMO channels. A two-dimensional ice-filling (2DIF) algorithm is proposed to efficiently accomplish this objective. The algorithm is motivated by the fact that the eigenspace of MIMO channel covariance can be decoupled into two sub-eigenspaces, which are associated with the correlations of transmitter antennas and receiver antennas, respectively. By properly setting the precoder and the combiner as the eigenvectors from these two sub-eigenspaces, the 2DIF promises to generate near-optimal observation matrices. Moreover, we further extend the 2DIF method to the popular hybrid combining systems, where a two-stage 2DIF (TS-2DIF) algorithm is developed to handle the analog combining circuits realized by phase shifters. Simulation results demonstrate that, compared to the state-of-the-art schemes, the proposed 2DIF and TS-2DIF methods can achieve superior channel estimation accuracy.

Problem

• Designing observation matrices to enhance densifying MIMO channel estimation
• Maximizing mutual information between received pilots and MIMO channels
• Extending the method to hybrid combining systems with phase shifters

Innovation

• Proposes 2D ice-filling algorithm for MIMO channel estimation
• Jointly designs combiners and precoders via mutual information
• Extends method to hybrid systems with phase shifters