Wonjin Sung
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Wonjin Sung has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
Jaewon Kim
Department of Electronic Engineering, Sogang University, Seoul, 121-742, Korea
4 shared publicationsJonghyun Park
Advanced Standard R&D Lab, LG Electronics, Seoul 07336, Republic of Korea
4 shared publications- Sangchun Park
Department of Electronic Engineering, Sogang University, Seoul, South Korea
4 shared publications Donghyun Lee
3 shared publicationsChanghyeon Kim
Department of Electronic Engineering Sogang University Seoul 04107 Republic of Korea
2 shared publicationsHyukjun Lee
Department of Electronic Engineering, Sogang University, Seoul 04107, Republic of Korea
2 shared publications
Latest Publications
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-14
AI/ML-Driven Operation of Large-Scale Dynamic Subarrays in Sub-6GHz under 5G NR Standards
To address the trade-offs between performance and hardware complexity in sub-6GHz massive multiple-input multiple-output (MIMO) systems, this paper proposes a deep learning (DL)-based approach for dynamic subarray configuration and mechanical downtilt optimization. The proposed method is tailored for sub-6GHz 5G systems, where constructing compact massive arrays is challenging and power consumption is a critical issue. In the proposed framework, both channel estimation and beamforming rely solely on the precoding matrix indicator (PMI), a 3GPP standardized feedback based on reference signaling. A supervised DL model dynamically configures subarrays and selects desirable array structures from a set of predefined configurations. This approach enables low-complexity and practical transceiver operations, with the advantage of potential applicability to commercial array hardware equipped with many transceiver units (TXRUs). The DL model is trained using diverse virtual environments generated under the 3GPP standard channel model. Performance evaluations in both virtual environments and digital twin-based scenarios show that the proposed method achieves a sum-rate comparable to that of exhaustive tilting and subarray selection, while providing a 13% gain over fixed configurations. Compared to a fully digital array, the method reduces the number of RF chains to one-sixth, incurring only a 7.1% sum-rate loss while achieving an energy efficiency gain of approximately 273%. These results, with the gains validated in digital twin-based scenarios, suggest that the proposed method can be well suited for real-world deployments.
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-14
AI/ML-Driven Operation of Large-Scale Dynamic Subarrays in Sub-6GHz under 5G NR Standards
To address the trade-offs between performance and hardware complexity in sub-6GHz massive multiple-input multiple-output (MIMO) systems, this paper proposes a deep learning (DL)-based approach for dynamic subarray configuration and mechanical downtilt optimization. The proposed method is tailored for sub-6GHz 5G systems, where constructing compact massive arrays is challenging and power consumption is a critical issue. In the proposed framework, both channel estimation and beamforming rely solely on the precoding matrix indicator (PMI), a 3GPP standardized feedback based on reference signaling. A supervised DL model dynamically configures subarrays and selects desirable array structures from a set of predefined configurations. This approach enables low-complexity and practical transceiver operations, with the advantage of potential applicability to commercial array hardware equipped with many transceiver units (TXRUs). The DL model is trained using diverse virtual environments generated under the 3GPP standard channel model. Performance evaluations in both virtual environments and digital twin-based scenarios show that the proposed method achieves a sum-rate comparable to that of exhaustive tilting and subarray selection, while providing a 13% gain over fixed configurations. Compared to a fully digital array, the method reduces the number of RF chains to one-sixth, incurring only a 7.1% sum-rate loss while achieving an energy efficiency gain of approximately 273%. These results, with the gains validated in digital twin-based scenarios, suggest that the proposed method can be well suited for real-world deployments.
- Open Access
- 9 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 22434-22444
Federated Learning for Beam Management in 5G and Beyond: A Collaborative AI/ML Approach
Building on the standardization efforts in 5G-Advanced for artificial intelligence (AI) and machine learning (ML)-based beam management (BM), two primary types of beam prediction models have been considered for deployment: the user equipment (UE)-sided model which is trained and deployed on the UE, and the network (NW)-sided model which is trained and deployed at the base station (BS). However, the NW-sided model suffers from significant communication overhead, while the UE-sided model is constrained by its reliance on training with data with limited diversity. To address these challenges, this paper proposes a federated learning (FL)-based approach as an alternative to centralized deployment at either the BS or the UE. The proposed FL-based model utilizes locally collected physical layer reference signal received power (L1-RSRP) data from each UE to train local models, with only the model parameters transmitted to the BS for aggregation. This approach achieves beam prediction performance comparable to that of the NW-sided model while reducing communication overhead by a factor of 4.3 and outperforming the UE-sided model by 11.2%.
- 9 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 22434-22444
Open AccessFederated Learning for Beam Management in 5G and Beyond: A Collaborative AI/ML Approach
Building on the standardization efforts in 5G-Advanced for artificial intelligence (AI) and machine learning (ML)-based beam management (BM), two primary types of beam prediction models have been considered for deployment: the user equipment (UE)-sided model which is trained and deployed on the UE, and the network (NW)-sided model which is trained and deployed at the base station (BS). However, the NW-sided model suffers from significant communication overhead, while the UE-sided model is constrained by its reliance on training with data with limited diversity. To address these challenges, this paper proposes a federated learning (FL)-based approach as an alternative to centralized deployment at either the BS or the UE. The proposed FL-based model utilizes locally collected physical layer reference signal received power (L1-RSRP) data from each UE to train local models, with only the model parameters transmitted to the BS for aggregation. This approach achieves beam prediction performance comparable to that of the NW-sided model while reducing communication overhead by a factor of 4.3 and outperforming the UE-sided model by 11.2%.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15, 3801-3805
DQN-Based Beam Tracking Using Reward-Aware Exploration Without Additional Signaling
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15, 3801-3805
DQN-Based Beam Tracking Using Reward-Aware Exploration Without Additional Signaling
- 11 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (2) , 549-553
Advanced Codebook Generation Using High-Resolution k-Means Clustering to PMI Data
An enhanced codebook generation approach based only on precoding matrix indicator (PMI) feedback information is proposed. By utilizing the kernel density estimation (KDE) to produce higher-resolution distributions from discrete PMI data and applying the k-means clustering algorithm, the approach iteratively updates the set of beamforming vectors to autonomously adapt to the channel characteristics and user equipment (UE) positions. The proposed codebook is 3GPP standard compliant in the sense that it can be applied to existing systems to achieve significantly enhanced beamforming performance, without any alterations to reference signaling or channel measurement report.
- 11 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (2) , 549-553
Advanced Codebook Generation Using High-Resolution k-Means Clustering to PMI Data
An enhanced codebook generation approach based only on precoding matrix indicator (PMI) feedback information is proposed. By utilizing the kernel density estimation (KDE) to produce higher-resolution distributions from discrete PMI data and applying the k-means clustering algorithm, the approach iteratively updates the set of beamforming vectors to autonomously adapt to the channel characteristics and user equipment (UE) positions. The proposed codebook is 3GPP standard compliant in the sense that it can be applied to existing systems to achieve significantly enhanced beamforming performance, without any alterations to reference signaling or channel measurement report.
- Open Access
- 18 July 2024
- Published by Springer Nature in EURASIP Journal on Wireless Communications and Networking
- Vol. 2024 (1) , 1-22
Digital twin based DDPG reinforcement learning for sum-rate maximization of AI-UAV communications
Construction of wireless infrastructure using unmanned aerial vehicle (UAV) can effectively expand the coverage and support high-density traffic of next-generation communication systems. Designing wireless systems including UAVs as aerial base stations (ABSs) is a challenging task, due to the mobility of ABSs causing time-varying nature of environmental surroundings and relative propagation paths to user equipment (UE) devices. Therefore, it is essential to have an accurate estimate of the channel for varying positioning of the UAVs. In this paper, we propose to adopt a digital twin based performance evaluation procedure for wireless systems including ABSs, providing enhanced accuracy of channel modeling for specific target deployment areas. Using ray-tracing channel models reflecting detailed building and terrain information of the transmission environment, an UAV position optimization algorithm based on reinforcement learning is presented. By utilizing deep deterministic policy gradient (DDPG), the proposed algorithm calculates the overall throughput in the digital twin and determines the desired states of the UAV. Performance evaluation results demonstrate the trajectory training ability of the algorithm and the performance advantage of the system with a reduced amount of shadow area compared to those with ground base stations (GBSs).
- 18 July 2024
- Published by Springer Nature in EURASIP Journal on Wireless Communications and Networking
- Vol. 2024 (1) , 1-22
Open AccessDigital twin based DDPG reinforcement learning for sum-rate maximization of AI-UAV communications
Construction of wireless infrastructure using unmanned aerial vehicle (UAV) can effectively expand the coverage and support high-density traffic of next-generation communication systems. Designing wireless systems including UAVs as aerial base stations (ABSs) is a challenging task, due to the mobility of ABSs causing time-varying nature of environmental surroundings and relative propagation paths to user equipment (UE) devices. Therefore, it is essential to have an accurate estimate of the channel for varying positioning of the UAVs. In this paper, we propose to adopt a digital twin based performance evaluation procedure for wireless systems including ABSs, providing enhanced accuracy of channel modeling for specific target deployment areas. Using ray-tracing channel models reflecting detailed building and terrain information of the transmission environment, an UAV position optimization algorithm based on reinforcement learning is presented. By utilizing deep deterministic policy gradient (DDPG), the proposed algorithm calculates the overall throughput in the digital twin and determines the desired states of the UAV. Performance evaluation results demonstrate the trajectory training ability of the algorithm and the performance advantage of the system with a reduced amount of shadow area compared to those with ground base stations (GBSs).
- Open Access
- 25 November 2022
- Published by Wiley in International Journal of Antennas and Propagation
- Vol. 2022, 1-12
A Next-Generation Codebook Evolution Strategy for Massive Arrays Using Deep Neurals Networks
The discrete Fourier transform (DFT)-based codebook is currently among the mostly commonly adopted codebooks for beamforming using arrays of different shapes and sizes, including the large-scale two-dimensional uniform planar array (UPA). DFT-based codevectors can be easily generated in arbitrary angle resolutions and apply well to millimeter-wave (mmWave) channels due to their directive nature of resulting beams. However, a fixed set of codevectors is applied regardless of the user distributions and the propagation environment, which may exhibit limited beamforming performance under certain transmission scenarios. In this paper, we propose a new way of generating a set of beamforming vectors for multiple-input multiple-output (MIMO) transmission using massive arrays under the limited feedback of the channel state information (CSI). Precoder matrix indicator (PMI) and channel quality indicator (CQI) reports from the users have become the sources for the generation of a new set of codevectors, which are autonomously determined by the deep learning (DL) module at the base station (BS). The process is operated in an iterative fashion to produce updated versions of the codebook with the reduced return of the loss function at the deep neural network (DNN). The time-varying codebook for each BS automatically reflects the characteristics of a given wireless environment to adapt to its channel and traffic conditions. The reference signal (RS) at the BS is periodically transmitted in the form of beamformed CSI-RS, thus the operation is transparent to the users of the system and no significant specification changes are necessary. A simple plug-and-play type of BS installation suffices to achieve the potential gain of the proposal, which is demonstrated by the implementation details of the DL engine and the corresponding performance simulation results.
- 25 November 2022
- Published by Wiley in International Journal of Antennas and Propagation
- Vol. 2022, 1-12
Open AccessA Next-Generation Codebook Evolution Strategy for Massive Arrays Using Deep Neurals Networks
The discrete Fourier transform (DFT)-based codebook is currently among the mostly commonly adopted codebooks for beamforming using arrays of different shapes and sizes, including the large-scale two-dimensional uniform planar array (UPA). DFT-based codevectors can be easily generated in arbitrary angle resolutions and apply well to millimeter-wave (mmWave) channels due to their directive nature of resulting beams. However, a fixed set of codevectors is applied regardless of the user distributions and the propagation environment, which may exhibit limited beamforming performance under certain transmission scenarios. In this paper, we propose a new way of generating a set of beamforming vectors for multiple-input multiple-output (MIMO) transmission using massive arrays under the limited feedback of the channel state information (CSI). Precoder matrix indicator (PMI) and channel quality indicator (CQI) reports from the users have become the sources for the generation of a new set of codevectors, which are autonomously determined by the deep learning (DL) module at the base station (BS). The process is operated in an iterative fashion to produce updated versions of the codebook with the reduced return of the loss function at the deep neural network (DNN). The time-varying codebook for each BS automatically reflects the characteristics of a given wireless environment to adapt to its channel and traffic conditions. The reference signal (RS) at the BS is periodically transmitted in the form of beamformed CSI-RS, thus the operation is transparent to the users of the system and no significant specification changes are necessary. A simple plug-and-play type of BS installation suffices to achieve the potential gain of the proposal, which is demonstrated by the implementation details of the DL engine and the corresponding performance simulation results.
- Open Access
- 3 May 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 47461-47475
Utilization of Beam Signatures Supporting High User Mobility With Extremely Low Feedback Overhead
Accurate beamforming under the constraint of limited-feedback for the channel state information (CSI) has always been a challenging task, despite its huge impact on the quality of multiple-input multiple-output (MIMO) transmission. The task is becoming especially important for millimeter-wave (mmWave) transmission which requires high-gain beams to overcome the severe pathloss experienced over the radio channel, since an inaccurate beam direction may cause a noticeable performance degradation. The signal blockage in the urban environment due to the mobile and human traffic can also degrade the beamforming performance, by generating blind spots for signal transmission as well as the CSI feedback. In this paper, a new way of transmitting accurate beams to highly mobile users with a substantially reduced amount of feedback overhead is proposed, by introducing a set of beam signatures that are composed of multiple beams along the trajectories of mobile users. Instead of forming a spot beam corresponding to the precoder matrix indicator (PMI) reported by the user equipment (UE), the base station (BS) utilizes the history of previous reports to determine an appropriate beam signature and transmit beams to predicted UE positions. The proactive decision for the next beam position is made with the aid of deep learning (DL) using the train data obtained from typical mobile movements for given road conditions, thus providing the adaptability to the channel environment with progressively improving accuracy. The set of beam signatures, which are called the beambook , includes the time dimension added to the conventional spatial dimension for beams to develop into a spatio-temporal codebook. The beambook produces enhanced and reliable beamforming over the mobile’s trajectory, even when the CSI feedback interval is considerably longer than parameters supported by the current 5G new radio (NR) standard. It is demonstrated that the proposed beambook significantly outperforms the conventional codebooks based on the discrete Fourier transform (DFT) matrix and the vector quantization (VQ) in both beamforming accuracy and throughput performance.
- 3 May 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 47461-47475
Open AccessUtilization of Beam Signatures Supporting High User Mobility With Extremely Low Feedback Overhead
Accurate beamforming under the constraint of limited-feedback for the channel state information (CSI) has always been a challenging task, despite its huge impact on the quality of multiple-input multiple-output (MIMO) transmission. The task is becoming especially important for millimeter-wave (mmWave) transmission which requires high-gain beams to overcome the severe pathloss experienced over the radio channel, since an inaccurate beam direction may cause a noticeable performance degradation. The signal blockage in the urban environment due to the mobile and human traffic can also degrade the beamforming performance, by generating blind spots for signal transmission as well as the CSI feedback. In this paper, a new way of transmitting accurate beams to highly mobile users with a substantially reduced amount of feedback overhead is proposed, by introducing a set of beam signatures that are composed of multiple beams along the trajectories of mobile users. Instead of forming a spot beam corresponding to the precoder matrix indicator (PMI) reported by the user equipment (UE), the base station (BS) utilizes the history of previous reports to determine an appropriate beam signature and transmit beams to predicted UE positions. The proactive decision for the next beam position is made with the aid of deep learning (DL) using the train data obtained from typical mobile movements for given road conditions, thus providing the adaptability to the channel environment with progressively improving accuracy. The set of beam signatures, which are called the beambook , includes the time dimension added to the conventional spatial dimension for beams to develop into a spatio-temporal codebook. The beambook produces enhanced and reliable beamforming over the mobile’s trajectory, even when the CSI feedback interval is considerably longer than parameters supported by the current 5G new radio (NR) standard. It is demonstrated that the proposed beambook significantly outperforms the conventional codebooks based on the discrete Fourier transform (DFT) matrix and the vector quantization (VQ) in both beamforming accuracy and throughput performance.
- Open Access
- 27 November 2021
- Published by Springer Nature in EURASIP Journal on Wireless Communications and Networking
- Vol. 2021 (1) , 1-18
Parametric codebook design for efficient signal transmission using uniform circular arrays
In order to improve the quality of the received signal and system spectral efficiency, accurate beamforming using a given antenna array is essential for multiple-input multiple-output (MIMO) systems. To obtain desired MIMO transmission performance, construction of codebooks which are composed of matching beamforming vectors to the array structure is important. To effectively cover different types of mobile traffic, the base station for 5G new radio employs antenna arrays in various sizes and shapes. Nevertheless, the codebooks adopted by the 3GPP standard so far are based on the uniform linear array and the uniform planar array, necessitating design techniques for a wider class of antenna arrays. In this paper, we propose codebook construction methods for the uniform circular array with parameters to flexibly set the initial phase and step size based on the channel characteristics of the user equipment (UE). When tested over the 3GPP spatial channel model, the proposed codebooks show a substantial amount of gain over the conventional codebooks in all UE locations within the cell.
- 27 November 2021
- Published by Springer Nature in EURASIP Journal on Wireless Communications and Networking
- Vol. 2021 (1) , 1-18
Open AccessParametric codebook design for efficient signal transmission using uniform circular arrays
In order to improve the quality of the received signal and system spectral efficiency, accurate beamforming using a given antenna array is essential for multiple-input multiple-output (MIMO) systems. To obtain desired MIMO transmission performance, construction of codebooks which are composed of matching beamforming vectors to the array structure is important. To effectively cover different types of mobile traffic, the base station for 5G new radio employs antenna arrays in various sizes and shapes. Nevertheless, the codebooks adopted by the 3GPP standard so far are based on the uniform linear array and the uniform planar array, necessitating design techniques for a wider class of antenna arrays. In this paper, we propose codebook construction methods for the uniform circular array with parameters to flexibly set the initial phase and step size based on the channel characteristics of the user equipment (UE). When tested over the 3GPP spatial channel model, the proposed codebooks show a substantial amount of gain over the conventional codebooks in all UE locations within the cell.
- Open Access
- 15 June 2021
- Published by Wiley in International Journal of Antennas and Propagation
- Vol. 2021, 1-9
2D-OPC Subarray Structure for Efficient Hybrid Beamforming over Sparse mmWave Channels
Millimeter-wave (mmWave) communication is a key technology of 5G new radio (NR) mobile communication systems. Efficient beamforming using a large antenna array is important to cope with the significant path loss experienced in the mmWave spectrum. The existing fully digital beamforming scheme requires a separate radio frequency (RF) chain for each antenna, which results in an excessive hardware cost and consumption power. Under these circumstances, hybrid beamforming which approaches the performance of fully digital beamforming while reducing the complexity is a promising solution for the mmWave multiuser transmission. By extending the existing hybrid beamforming strategies, this paper proposes a novel architecture which effectively reduces the hardware cost and complexity for large antenna arrays. The proposed scheme includes multiple subarrays in the form of uniform planar array (UPA) which are allowed to be overlapped in the two-dimensional space. The corresponding antenna structure is referred to as the two-dimensional overlapped partially connected (2D-OPC) subarray structure. We evaluate the performance of the proposed scheme to suggest performance-complexity trade-offs in designing versatile antenna arrays for efficient beamforming over the mmWave channel.
- 15 June 2021
- Published by Wiley in International Journal of Antennas and Propagation
- Vol. 2021, 1-9
Open Access2D-OPC Subarray Structure for Efficient Hybrid Beamforming over Sparse mmWave Channels
Millimeter-wave (mmWave) communication is a key technology of 5G new radio (NR) mobile communication systems. Efficient beamforming using a large antenna array is important to cope with the significant path loss experienced in the mmWave spectrum. The existing fully digital beamforming scheme requires a separate radio frequency (RF) chain for each antenna, which results in an excessive hardware cost and consumption power. Under these circumstances, hybrid beamforming which approaches the performance of fully digital beamforming while reducing the complexity is a promising solution for the mmWave multiuser transmission. By extending the existing hybrid beamforming strategies, this paper proposes a novel architecture which effectively reduces the hardware cost and complexity for large antenna arrays. The proposed scheme includes multiple subarrays in the form of uniform planar array (UPA) which are allowed to be overlapped in the two-dimensional space. The corresponding antenna structure is referred to as the two-dimensional overlapped partially connected (2D-OPC) subarray structure. We evaluate the performance of the proposed scheme to suggest performance-complexity trade-offs in designing versatile antenna arrays for efficient beamforming over the mmWave channel.