Yunfeng Guan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yunfeng Guan has a total of 75 co-authors. Here are the top co-authors by number of shared publications:
- Yin Xu
Shanghai Jiao Tong University
17 shared publications - Dazhi He
Shanghai Jiao Tong University
16 shared publications - Wenjun Zhang
Shanghai Jiao Tong University
13 shared publications - Yihang Huang
Chongqing University of Posts and Telecommunications
7 shared publications - Meixia Tao
Shanghai Jiao Tong University
6 shared publications - Haoyang Li
Cooperative Medianet Innovation Center (CMIC), Shanghai Jiao Tong University, Shanghai, China
4 shared publications
Latest Publications
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32  (7) , 6363-6378
Deformable 2D Gaussian Splatting for Efficient Wireless Radiance Field Rendering
Modeling the wireless radiance field (WRF) is fundamental to modern communication systems, enabling key tasks such as localization, sensing, and channel estimation. Traditional approaches, which rely on empirical formulas or physical simulations, often suffer from limited accuracy or require strong scene priors. Recent neural radiance field (NeRF)-based methods improve reconstruction fidelity through differentiable volumetric rendering, but their reliance on computationally expensive multilayer perceptron (MLP) queries hinders real-time deployment. To overcome these challenges, we introduce Gaussian splatting (GS) to the wireless domain, leveraging its efficiency in modeling optical radiance fields to enable compact and accurate WRF reconstruction. Specifically, we propose SwiftWRF, a deformable 2D Gaussian splatting framework that synthesizes WRF spectra at arbitrary positions under single-sided transceiver mobility. SwiftWRF employs CUDA-accelerated rasterization to render spectra at over 100 k FPS and uses the lightweight MLP to model the deformation of 2D Gaussians, effectively capturing mobility-induced WRF variations. In addition to novel spectrum synthesis, the efficacy of SwiftWRF is further underscored in its applications in angle-of-arrival (AoA) and received signal strength indicator (RSSI) prediction. Experiments conducted on both real-world and synthetic indoor scenes demonstrate that SwiftWRF can reconstruct WRF spectra up to 500x faster than existing state-of-the-art methods, while significantly enhancing its signal quality.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 32  (7) , 6363-6378
Deformable 2D Gaussian Splatting for Efficient Wireless Radiance Field Rendering
Modeling the wireless radiance field (WRF) is fundamental to modern communication systems, enabling key tasks such as localization, sensing, and channel estimation. Traditional approaches, which rely on empirical formulas or physical simulations, often suffer from limited accuracy or require strong scene priors. Recent neural radiance field (NeRF)-based methods improve reconstruction fidelity through differentiable volumetric rendering, but their reliance on computationally expensive multilayer perceptron (MLP) queries hinders real-time deployment. To overcome these challenges, we introduce Gaussian splatting (GS) to the wireless domain, leveraging its efficiency in modeling optical radiance fields to enable compact and accurate WRF reconstruction. Specifically, we propose SwiftWRF, a deformable 2D Gaussian splatting framework that synthesizes WRF spectra at arbitrary positions under single-sided transceiver mobility. SwiftWRF employs CUDA-accelerated rasterization to render spectra at over 100 k FPS and uses the lightweight MLP to model the deformation of 2D Gaussians, effectively capturing mobility-induced WRF variations. In addition to novel spectrum synthesis, the efficacy of SwiftWRF is further underscored in its applications in angle-of-arrival (AoA) and received signal strength indicator (RSSI) prediction. Experiments conducted on both real-world and synthetic indoor scenes demonstrate that SwiftWRF can reconstruct WRF spectra up to 500x faster than existing state-of-the-art methods, while significantly enhancing its signal quality.
- Top Cited Papers
- 22 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36  (6) , 7992-8005
Light4GS: Lightweight Compact 4D Gaussian Splatting Generation via Context Model
3D Gaussian Splatting (3DGS) has emerged as an efficient and high-fidelity paradigm for novel view synthesis. To adapt 3DGS for dynamic content, deformable 3DGS incorporates temporally deformable primitives with learnable latent embeddings to capture complex motions. Despite its impressive performance, the high-dimensional embeddings and vast number of primitives lead to substantial storage requirements. In this paper, we introduce a Lightweight 4DGS framework, called Light4GS, that employs significance pruning with a deep context model to provide a lightweight storage-efficient dynamic 3DGS representation. The proposed Light4GS is based on 4DGS, which is a typical representation of deformable 3DGS. Specifically, our framework is built upon two core components: (1) a spatio-temporal significance pruning strategy that eliminates over 64% of the deformable primitives, followed by an entropy-constrained spherical harmonics compression applied to the remainder; and (2) a deep context model that integrates intra- and inter-prediction with hyperprior into a coarse-to-fine context structure, enabling efficient multiscale latent embedding compression. Our approach achieves over 12× compression and increases rendering FPS up to 20% compared to the baseline 4DGS, and also superior to frame-wise state-of-the-art 3DGS compression methods. Experiment results show the effectiveness of our Light4GS in terms of both intra- and inter-prediction methods without sacrificing rendering quality. The code is available at https://github.com/Evan-sudo/Light4GS.
- 22 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36  (6) , 7992-8005
Top Cited PapersLight4GS: Lightweight Compact 4D Gaussian Splatting Generation via Context Model
3D Gaussian Splatting (3DGS) has emerged as an efficient and high-fidelity paradigm for novel view synthesis. To adapt 3DGS for dynamic content, deformable 3DGS incorporates temporally deformable primitives with learnable latent embeddings to capture complex motions. Despite its impressive performance, the high-dimensional embeddings and vast number of primitives lead to substantial storage requirements. In this paper, we introduce a Lightweight 4DGS framework, called Light4GS, that employs significance pruning with a deep context model to provide a lightweight storage-efficient dynamic 3DGS representation. The proposed Light4GS is based on 4DGS, which is a typical representation of deformable 3DGS. Specifically, our framework is built upon two core components: (1) a spatio-temporal significance pruning strategy that eliminates over 64% of the deformable primitives, followed by an entropy-constrained spherical harmonics compression applied to the remainder; and (2) a deep context model that integrates intra- and inter-prediction with hyperprior into a coarse-to-fine context structure, enabling efficient multiscale latent embedding compression. Our approach achieves over 12× compression and increases rendering FPS up to 20% compared to the baseline 4DGS, and also superior to frame-wise state-of-the-art 3DGS compression methods. Experiment results show the effectiveness of our Light4GS in terms of both intra- and inter-prediction methods without sacrificing rendering quality. The code is available at https://github.com/Evan-sudo/Light4GS.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15, 5790-5794
SysUM: NDT-Assisted System-Level Scenario-Adaptive Universal Model for AI-6G Networks
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15, 5790-5794
SysUM: NDT-Assisted System-Level Scenario-Adaptive Universal Model for AI-6G Networks
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- p. 1-14
Adaptive Cross-Layer Video Delivery With StreamOptix
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- p. 1-14
Adaptive Cross-Layer Video Delivery With StreamOptix
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 10, 881-895
RIS-Assisted Low Complexity Precoding for Decentralized Massive MU-MIMO System
Massive multi-user multiple-input multiple-output (MU-MIMO) antenna arrays are pivotal for the advancement of next-generation communication system. However, under non-line-of-sight (NLoS) conditions, the performance improvements provided by Massive MU-MIMO are limited, and the complexity associated with precoding design approaches becomes overwhelming, particularly in scenarios involving multiuser variations. The advent of reconfigurable intelligent surfaces (RIS) offers a promising solution, enabling substantial performance enhancements under NLoS conditions while maintaining a low cost. Nevertheless, the complexity of precoding design schemes in RIS-assisted system with large-scale antenna arrays remains a significant challenge. In response to this issue, this article introduces the star-decentralized minimum mean square error Tomlinson-Harashima precoding (sDMMSE-THP) algorithm and integrates it into the star decentralized baseband processing (DBP) framework. This combination not only reduces the computational complexity but also delivers superior system performance. To further mitigate the issue of increasing precoding update complexity due to user variations, we propose an update algorithm for sDMMSE-THP that avoids redundant computations arising from the number of user changes, thus ensuring that the overall system complexity remains manageable. Simulation results demonstrate that the proposed DBP approach for precoding in RIS-assisted system, particularly in large-scale antenna array scenarios, effectively maintains high performance while significantly reducing system complexity. These findings underscore the importance and potential of the proposed methodology in enhancing the efficiency of future communication system.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 10, 881-895
RIS-Assisted Low Complexity Precoding for Decentralized Massive MU-MIMO System
Massive multi-user multiple-input multiple-output (MU-MIMO) antenna arrays are pivotal for the advancement of next-generation communication system. However, under non-line-of-sight (NLoS) conditions, the performance improvements provided by Massive MU-MIMO are limited, and the complexity associated with precoding design approaches becomes overwhelming, particularly in scenarios involving multiuser variations. The advent of reconfigurable intelligent surfaces (RIS) offers a promising solution, enabling substantial performance enhancements under NLoS conditions while maintaining a low cost. Nevertheless, the complexity of precoding design schemes in RIS-assisted system with large-scale antenna arrays remains a significant challenge. In response to this issue, this article introduces the star-decentralized minimum mean square error Tomlinson-Harashima precoding (sDMMSE-THP) algorithm and integrates it into the star decentralized baseband processing (DBP) framework. This combination not only reduces the computational complexity but also delivers superior system performance. To further mitigate the issue of increasing precoding update complexity due to user variations, we propose an update algorithm for sDMMSE-THP that avoids redundant computations arising from the number of user changes, thus ensuring that the overall system complexity remains manageable. Simulation results demonstrate that the proposed DBP approach for precoding in RIS-assisted system, particularly in large-scale antenna array scenarios, effectively maintains high performance while significantly reducing system complexity. These findings underscore the importance and potential of the proposed methodology in enhancing the efficiency of future communication system.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14  (8) , 2651-2655
AI2MMUM: AI-AI Oriented Multi-Modal Universal Model Leveraging Telecom Domain Large Model
Designing a 6G-oriented universal model capable of processing multi-modal data and executing diverse air interface tasks has emerged as a common goal in future wireless systems. Building on our prior work in communication multi-modal alignment and telecom large language model (LLM), we propose a scalable, task-aware artificial intelligence-air interface multi-modal universal model (AIMMUM), which flexibility and effectively perform various physical layer tasks according to subtle task instructions. The LLM backbone provides robust contextual comprehension and generalization capabilities, while a fine-tuning approach is adopted to incorporate domain-specific knowledge. To enhance task adaptability, task instructions consist of fixed task keywords and learnable, implicit prefix prompts. Frozen radio modality encoders extract universal representations and adapter layers subsequently bridge radio and language modalities. Moreover, lightweight task-specific heads are designed to directly output task objectives. Comprehensive evaluations demonstrate that AIMMUM achieves SOTA performance across five representative physical environment/wireless channel-based downstream tasks using the WAIR-D and DeepMIMO datasets.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14  (8) , 2651-2655
AI2MMUM: AI-AI Oriented Multi-Modal Universal Model Leveraging Telecom Domain Large Model
Designing a 6G-oriented universal model capable of processing multi-modal data and executing diverse air interface tasks has emerged as a common goal in future wireless systems. Building on our prior work in communication multi-modal alignment and telecom large language model (LLM), we propose a scalable, task-aware artificial intelligence-air interface multi-modal universal model (AIMMUM), which flexibility and effectively perform various physical layer tasks according to subtle task instructions. The LLM backbone provides robust contextual comprehension and generalization capabilities, while a fine-tuning approach is adopted to incorporate domain-specific knowledge. To enhance task adaptability, task instructions consist of fixed task keywords and learnable, implicit prefix prompts. Frozen radio modality encoders extract universal representations and adapter layers subsequently bridge radio and language modalities. Moreover, lightweight task-specific heads are designed to directly output task objectives. Comprehensive evaluations demonstrate that AIMMUM achieves SOTA performance across five representative physical environment/wireless channel-based downstream tasks using the WAIR-D and DeepMIMO datasets.
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 24  (9) , 7377-7391
Addressing the Curse of Scenario and Task Generalization in AI-6G: A Multi-Modal Paradigm
Existing works on machine learning (ML)-empowered wireless communication primarily focus on monolithic scenarios and single tasks. However, with the blooming growth of communication task classes coupled with various task requirements in future 6G systems, this working pattern is obviously unsustainable. Therefore, identifying a groundbreaking paradigm that enables a universal model to solve multiple tasks in the physical layer within diverse scenarios is crucial for future system evolution. This paper aims to fundamentally address the curse of ML model generalization across diverse scenarios and tasks by unleashing multi-modal feature integration capabilities in future systems. Given the universality of electromagnetic propagation theory, the communication process is determined by the scattering environment, which can be more comprehensively characterized by cross-modal perception, thus providing sufficient information for all communication tasks across varied environments. This fact motivates us to propose a transformative two-stage multi-modal pre-training and downstream task adaptation paradigm. In the pre-training stage, we introduce a multi-modal two-tower model and a corresponding contrastive learning method to integrate the explicit description of the scattering environment and implicit channel state information (CSI) into a universal representation, which encapsulates rich high-level knowledge and can be leveraged for all downstream tasks in different scenarios. Additionally, we present two specially designed model structures to enhance the interaction of communication modalities. In the second stage, based on the frozen pre-trained model, we propose a direct method and a pluggable method for flexible and low-cost task adaptation. Experimental results demonstrate that our proposed approach significantly outperforms benchmarks in both task performance and tuning parameter size for exemplary sub-tasks in unseen scenarios.
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 24  (9) , 7377-7391
Addressing the Curse of Scenario and Task Generalization in AI-6G: A Multi-Modal Paradigm
Existing works on machine learning (ML)-empowered wireless communication primarily focus on monolithic scenarios and single tasks. However, with the blooming growth of communication task classes coupled with various task requirements in future 6G systems, this working pattern is obviously unsustainable. Therefore, identifying a groundbreaking paradigm that enables a universal model to solve multiple tasks in the physical layer within diverse scenarios is crucial for future system evolution. This paper aims to fundamentally address the curse of ML model generalization across diverse scenarios and tasks by unleashing multi-modal feature integration capabilities in future systems. Given the universality of electromagnetic propagation theory, the communication process is determined by the scattering environment, which can be more comprehensively characterized by cross-modal perception, thus providing sufficient information for all communication tasks across varied environments. This fact motivates us to propose a transformative two-stage multi-modal pre-training and downstream task adaptation paradigm. In the pre-training stage, we introduce a multi-modal two-tower model and a corresponding contrastive learning method to integrate the explicit description of the scattering environment and implicit channel state information (CSI) into a universal representation, which encapsulates rich high-level knowledge and can be leveraged for all downstream tasks in different scenarios. Additionally, we present two specially designed model structures to enhance the interaction of communication modalities. In the second stage, based on the frozen pre-trained model, we propose a direct method and a pluggable method for flexible and low-cost task adaptation. Experimental results demonstrate that our proposed approach significantly outperforms benchmarks in both task performance and tuning parameter size for exemplary sub-tasks in unseen scenarios.
- 14 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- Vol. 71  (2) , 426-438
Gray-Mapped NOM-Enhanced SFN: A Broadcast and Broadband Converged Transmission Solution in LTE-Based 5G Broadcast
Broadcast and broadband converged transmission has emerged as a prominent research focus within broadcast technology. Abundant corresponding studies have been conducted in traditional terrestrial broadcast and 3GPP unicast systems. However, due to issues like system compatibility, traditional terrestrial broadcasts usually reveal insufficient flexibility in transmitting broadband services, and conventional unicast systems always perform inefficiently in delivering broadcast services in scenarios of converged transmission. In addition, as the current Non-Orthogonal Multiplexing (NOM) scheme employed in converged transmission usually does not comply with the Gray-mapping rule, the required codeword-level Successive Interference Cancellation (SIC) algorithm makes the Enhanced Layer (EL) data share the same processing delay as the Core Layer (CL) one, which restricts the variety of EL services. This paper focuses on the physical layer technologies of converged transmission in the 3GPP LTE-based 5G Broadcast system. Due to the inherent good compatibility with both broadcast and broadband systems, LTE-based 5G Broadcast has great potential in realizing the converged transmission of broadcast and broadband. In addition, a novel converged transmission scheme enhanced by Gray-mapped NOM is proposed in this paper, and the corresponding networking architecture, frame structure, transmitting processing, and receiving algorithms are put forward. By significantly improving the performance of the non-SIC receiving algorithm, the proposed Gray-mapped NOM-enhanced SFN (GNeSFN) scheme enables the EL customized services and the CL broadcast services to have processing delays independent from each other, bringing more flexibility to converged transmission. Link-level simulations are carried out with different system configurations and multiple channel scenarios, verifying the effectiveness and feasibility of the proposed scheme.
- 14 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- Vol. 71  (2) , 426-438
Gray-Mapped NOM-Enhanced SFN: A Broadcast and Broadband Converged Transmission Solution in LTE-Based 5G Broadcast
Broadcast and broadband converged transmission has emerged as a prominent research focus within broadcast technology. Abundant corresponding studies have been conducted in traditional terrestrial broadcast and 3GPP unicast systems. However, due to issues like system compatibility, traditional terrestrial broadcasts usually reveal insufficient flexibility in transmitting broadband services, and conventional unicast systems always perform inefficiently in delivering broadcast services in scenarios of converged transmission. In addition, as the current Non-Orthogonal Multiplexing (NOM) scheme employed in converged transmission usually does not comply with the Gray-mapping rule, the required codeword-level Successive Interference Cancellation (SIC) algorithm makes the Enhanced Layer (EL) data share the same processing delay as the Core Layer (CL) one, which restricts the variety of EL services. This paper focuses on the physical layer technologies of converged transmission in the 3GPP LTE-based 5G Broadcast system. Due to the inherent good compatibility with both broadcast and broadband systems, LTE-based 5G Broadcast has great potential in realizing the converged transmission of broadcast and broadband. In addition, a novel converged transmission scheme enhanced by Gray-mapped NOM is proposed in this paper, and the corresponding networking architecture, frame structure, transmitting processing, and receiving algorithms are put forward. By significantly improving the performance of the non-SIC receiving algorithm, the proposed Gray-mapped NOM-enhanced SFN (GNeSFN) scheme enables the EL customized services and the CL broadcast services to have processing delays independent from each other, bringing more flexibility to converged transmission. Link-level simulations are carried out with different system configurations and multiple channel scenarios, verifying the effectiveness and feasibility of the proposed scheme.
- Open Access
Addressing the Curse of Scenario and Task Generalization in AI-6G: A Multi-Modal Paradigm
Existing works on machine learning (ML)-empowered wireless communication primarily focus on monolithic scenarios and single tasks. However, with the blooming growth of communication task classes coupled with various task requirements in future 6G systems, this working pattern is obviously unsustainable. Therefore, identifying a groundbreaking paradigm that enables a universal model to solve multiple tasks in the physical layer within diverse scenarios is crucial for future system evolution. This paper aims to fundamentally address the curse of ML model generalization across diverse scenarios and tasks by unleashing multi-modal feature integration capabilities in future systems. Given the universality of electromagnetic propagation theory, the communication process is determined by the scattering environment, which can be more comprehensively characterized by cross-modal perception, thus providing sufficient information for all communication tasks across varied environments. This fact motivates us to propose a transformative two-stage multi-modal pre-training and downstream task adaptation paradigm...
Open AccessAddressing the Curse of Scenario and Task Generalization in AI-6G: A Multi-Modal Paradigm
Existing works on machine learning (ML)-empowered wireless communication primarily focus on monolithic scenarios and single tasks. However, with the blooming growth of communication task classes coupled with various task requirements in future 6G systems, this working pattern is obviously unsustainable. Therefore, identifying a groundbreaking paradigm that enables a universal model to solve multiple tasks in the physical layer within diverse scenarios is crucial for future system evolution. This paper aims to fundamentally address the curse of ML model generalization across diverse scenarios and tasks by unleashing multi-modal feature integration capabilities in future systems. Given the universality of electromagnetic propagation theory, the communication process is determined by the scattering environment, which can be more comprehensively characterized by cross-modal perception, thus providing sufficient information for all communication tasks across varied environments. This fact motivates us to propose a transformative two-stage multi-modal pre-training and downstream task adaptation paradigm...
- Open Access
Efficient Precoding in XL-MIMO-AFDM System
This paper explores the potential of affine frequency division multiplexing (AFDM) to mitigate the multiuser interference (MUI) problem by employing time-domain precoding in extremely-large-scale multiple-input multiple-output (XL-MIMO) systems. In XL-MIMO systems, user mobility significantly improves network capacity and transmission quality. Meanwhile, the robustness of AFDM to Doppler shift is enhanced in user mobility scenarios, which further improves the system performance. However, the multicarrier nature of AFDM has attracted much attention, and it leads to a significant increase in precoding complexity. However, the serious problem is that the multicarrier use of AFDM leads to a sharp increase in precoding complexity. Therefore, we employ efficient precoding randomized Kaczmarz (rKA) to reduce the complexity overhead. Through simulation analysis, we compare the performance of XL-MIMO-AFDM and XL-MIMO orthogonal frequency division multiplexing (XL-MIMO-OFDM) in mobile scenarios, and the results show that our proposed AFDM-based XL-MIMO precoding design can be more efficient.
Open AccessEfficient Precoding in XL-MIMO-AFDM System
This paper explores the potential of affine frequency division multiplexing (AFDM) to mitigate the multiuser interference (MUI) problem by employing time-domain precoding in extremely-large-scale multiple-input multiple-output (XL-MIMO) systems. In XL-MIMO systems, user mobility significantly improves network capacity and transmission quality. Meanwhile, the robustness of AFDM to Doppler shift is enhanced in user mobility scenarios, which further improves the system performance. However, the multicarrier nature of AFDM has attracted much attention, and it leads to a significant increase in precoding complexity. However, the serious problem is that the multicarrier use of AFDM leads to a sharp increase in precoding complexity. Therefore, we employ efficient precoding randomized Kaczmarz (rKA) to reduce the complexity overhead. Through simulation analysis, we compare the performance of XL-MIMO-AFDM and XL-MIMO orthogonal frequency division multiplexing (XL-MIMO-OFDM) in mobile scenarios, and the results show that our proposed AFDM-based XL-MIMO precoding design can be more efficient.