Research

Some recent research results in the MIMO+X group

Attention-aided MMSE for OFDM Channel Estimation

Accurate channel estimation is essential for reliable OFDM communication, yet conventional MMSE methods rely on hard-to-obtain channel statistics, and deep learning models often suffer from high complexity and low interpretability.
To address this, we propose Attention-aided MMSE (A-MMSE), a model-based learning approach that leverages an Attention Transformer to learn a linear estimation filter. Once trained, the filter enables fast and efficient channel estimation using only linear operations. A two-stage attention encoder explicitly captures frequency and temporal correlations in the channel, leading to significantly improved accuracy and robustness. A rank-adaptive extension allows flexible control over complexity, making it suitable for real-world 5G and upcoming 6G systems.

Related Publications

T. Ha, C. Jung, H. Kim, J. Park, and J. Park, “Learning MMSE Filters for OFDM Channel Estimation: Attention Transformer Gains at Linear Inference,” submitted to IEEE Transactions on Wireless Communications. [Arxiv] [Github] (Under Revision)
T. Ha, H. Kim, and J. Park, “Attention-Aided MMSE with Ridge Denoiser: How to Train under Noisy Channel Samples,” submitted to IEEE Transactions on Vehicular Technology. (Under Revision)
T. Ha and J. Park, “Attention-Aided MMSE for OFDM Channel Estimation,” 2025 Asilomar Conference on Signals, Systems, and Computers (Asilomar).

Massive MIMO for Multibeam Satellites

Massive MIMO is a key enabler for next-generation satellite networks, offering high spectral efficiency and spatial multiplexing gains over wide coverage areas. Yet satellite channels differ fundamentally from terrestrial ones: they are dominated by line-of-sight (LoS) propagation, shaped by 3D geometry with uniform planar arrays (UPAs), and subject to large Doppler shifts arising from fast satellite motion. These features call for a rethinking of precoding, user scheduling, and beamforming principles that were originally developed for rich-scattering terrestrial systems.

Our group studies massive MIMO for satellite communications from both an analytical and an algorithmic perspective. On the analysis side, we characterize the asymptotic rate behavior of multibeam and multicast systems, revisiting multi-user diversity under LoS channels and 3D array geometry, and deriving how user density must scale with the number of antennas and beams to preserve asymptotic optimality. On the design side, we develop space-time beamforming methods that exploit the Doppler dimension as an additional signaling resource, enabling extremely narrow beams and robust transmission under spatial user crowding.

Related Publications

S. Kim, J. Choi, W. Shin, N. Lee, and J. Park, “Multibeam Satellite Communications with Massive MIMO: Asymptotic Performance Analysis and Design Insights,” in IEEE Transactions on Wireless Communications, vol. 24, no. 11, pp. 9449-9464, Nov. 2025. [Arxiv]
S. Kim and J. Park, “Asymptotic Scaling Law Analysis of Multicast Satellite Communications With Massive MIMO,” in IEEE Wireless Communications Letters, vol. 14, no. 12, pp. 4092-4096, Dec. 2025. [Arxiv]
J. Yim, J. Choi, J. Park, I. P. Roberts, and N. Lee, “Space-Time Beamforming for LEO Satellite Communications: Enabling Extremely Narrow Beams,” in IEEE Transactions on Wireless Communications, vol. 25, pp. 12725-12739, 2026. [Arxiv]
H. Yun, S. Kim, and J. Park, “Space-Time Adaptive Beamforming for Satellite Communications: Harnessing Doppler as New Signaling Dimensions,” submitted to IEEE Transactions on Wireless Communications. [Arxiv] (Under Revision)
J. Lim, J. Park, and N. Lee, “Space-Time Beamforming for Satellite Communications,” 2024 Asilomar Conference on Signals, Systems, and Computers (Asilomar).
H. Yun and J. Park, “Space-Time Adaptive Beamforming for Satellite Communications under Spatial User Crowding,” accepted to 2026 Asilomar Conference on Signals, Systems, and Computers (Asilomar).

Greedy RF Chain Selection for MIMO ISAC

In 6G systems, the role of MIMO ISAC (Integrated Sensing and Communication) is becoming increasingly prominent. The growth in antenna count necessitates a proportional increase in RF chains, resulting in substantially higher power consumption. To mitigate this, antenna (RF chain) selection has emerged as a key technique for reducing hardware cost and energy consumption in MIMO-ISAC systems.
In this work, we present a greedy RF chain selection for MIMO ISAC, in which candidate RF chains are sequentially selected based on their contribution to a unified mutual information (MI) performance metric.

Related Publications

S. Shin, S. Jung, J. Choi, and J. Park, “Efficient RF Chain Selection for MIMO Integrated Sensing and Communications: A Greedy Approach,” submitted to IEEE Transactions on Communications. [Arxiv]
S. Shin and J. Park, “Beam Selection in MIMO-ISAC Systems: A Greedy Approach,” accepted to 2025 Asilomar Conference on Signals, Systems, and Computers (Asilomar).

Shot in the Dark – MIMO without CSIT

In MIMO downlink systems, the acquisition of accurate CSIT has long been a major challenge, especially in scenarios with limited feedback or rapidly changing channels.
In our recent work, we propose a novel framework that avoids reliance on explicit CSIT feedback by leveraging partial channel reciprocity between the uplink and downlink. By exploiting shared channel characteristics, the proposed method directly predicts the downlink CSI from the uplink estimated CSI, enabling efficient and robust downlink transmission with significantly reduced overhead associated with CSI.

Related Publications

N. Kim, I. P. Roberts, and J. Park, “Splitting Messages in the Dark – Rate-Splitting Multiple Access for FDD Massive MIMO without CSI Feedback,” in IEEE Transactions on Wireless Communications, vol. 24, no. 4, pp. 3320-3332, April 2025. [Arxiv]
N. Kim, J. Han, J. Choi, A. Alkhateeb, C. -B. Chae, and J. Park, “Integrated Sensing and Communications in Downlink FDD MIMO without CSI Feedback,” in IEEE Transactions on Wireless Communications, vol. 25, pp. 2984-3000, 2026. [Arxiv]
J. Han, N. Kim, and J. Park, “Reducing Latency by Eliminating CSIT Feedback: FDD Downlink MIMO Transmission for Internet-of-Things Communications,” in IEEE Internet of Things Journal, vol. 13, no. 6, pp. 11301-11315, 15 March, 2026. [Arxiv]
J. Han and J. Park, “Feedback-Free Precoding for Low-Latency FDD Downlink MIMO for IoT Communications,” accepted to 2026 IEEE International Conference on Communications (ICC).
N. Kim and J. Park, “Robust Precoding and Rate-Splitting Multiple Access for FDD Massive MIMO Without CSI Feedback,” 2025 IEEE International Symposium on Information Theory (ISIT).

Y-Twin (made with Sionna)

We have developed a custom digital twin, named Y-Twin, which fully reflects the geometry, traffic dynamics, and mobility patterns of the Sinchon area near Yonsei University. Built on top of the Sionna framework, Y-Twin supports fully functioning RAN emulation, enabling end-to-end wireless system evaluation in a realistic urban environment. This platform is expected to serve as a foundation for a wide range of AI-native RAN research in the near future.
Dataset, structure, and usage details will be made available soon.

Related Publications

G. Kim and J. Park, “Y-Twin: A Digital-Twin Framework for Advanced Physical-Layer Research,” in preparation.

State Estimation with 1-bit Observation and Imperfect Model

State estimation is an important problem for estimating a system’s hidden state, and Kalman filter is commonly used for this purpose. Most of the existing algorithms assume ideal environment (infinite resolution or full model knowledge, etc), so their performance degrades in real-world environments.

This work considers a state estimation problem with 1-bit observation and imperfect models. By using the well-knonw Bussgang theorem to approximate nonlinear inputs linearly, we get a linear filter structure while accounting for ADC quantization distortions. To overcome model uncertainties and nonlinearities, we propose a hybrid DNN approach that integrates Bussgang-based linearization with data-driven learning for robust and efficient inference.

Related Publications

G. Choi, J. Park, N. Shlezinger, Y. C. Eldar, and N. Lee, “Split-KalmanNet: A Robust Model-based Deep Learning Approach for State Estimation,” in IEEE Transactions on Vehicular Technology, vol. 72, no. 9, pp. 12326-12331, Sept. 2023. [Arxiv].
C. Jung, T. Ha, H. Kim, and J. Park, “Energy-Efficient State Estimation with 1-Bit Sensing: A Bussgang-Kalman Framework for Internet-of-Things,” submitted to IEEE Internet of Things Journal. [Arxiv] [Github] (Under Revision)
C. Jung and J. Park, “Finding a Path from 1-bit Quantizer: State Estimation with 1-Bit Observations and Imperfect Models,” 2025 Asilomar Conference on Signals, Systems, and Computers (Asilomar).

MIMO-ME-MS Channel

In this work, we introduce the MIMO-ME-MS channel, which comprises a multi-antenna transmitter and receiver, along with a multi-antenna eavesdropper and sensing receiver. Within this framework, we investigate the quasi-optimal precoder structure via subspace analysis and propose an iterative precoding design method tailored to this setting.

Related Publications

S. Jung, N. Lee, and J. Park, “The MIMO-ME-MS Channel: Analysis and Algorithm for Secure MIMO Integrated Sensing and Communications,” in IEEE Journal of Selected Areas in Communications, vol. 44, pp. 4067-4082, 2026. [Arxiv]
S. Jung and J. Park, “The MIMO-ME-MS Channel: Useful-Subspace Characterization and Practical Precoder Design for Secure MIMO ISAC,” accepted to 2026 IEEE International Symposium on Information Theory (ISIT).

Research Support

Our group has been funded by a diverse range of government and industry sponsors. Our group gratefully acknowledges ongoing and past support from NRF, IITP, KRIT, Samsung, Intellian Technologies.