Fengxia Han
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Fengxia Han has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Shengjie Zhao
Tongji University, Shanghai, China
13 shared publications - Hao Deng
Tongji University
6 shared publications - Hao Jiang
Southeast University
3 shared publications - Xiaodong Wang
Columbia University
3 shared publications - Yuan Liu
School of Information Science and Electrical Engineering, Shandong Jiaotong University, Jinan 250357, China
3 shared publications - Hongming Zhang
School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
2 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (21) , 31758-31768
Adaptive and Redundancy-Aware Multimodal Fusion for Sensing-Assisted Beam Prediction
As wireless systems evolve toward higher frequency bands with narrower beams and increased mobility, ensuring accurate beam prediction is crucial for optimizing network performance and improving signal transmission efficiency. Recently, sensing-assisted communication has gained attention for leveraging environmental sensing data to enhance wireless decision-making. However, existing works mainly focus on incorporating more modalities while neglecting the negative impact of modality redundancy. In practice, certain modalities may contribute little to target tasks and introduce unnecessary computational overhead. To address such issue, we will investigate the representational effectiveness of diverse sensing modalities. Driven by numerical observations, we design a novel model architecture, termed as LRCT-MS, that integrates low-rank fusion mechanism with cross-modal transformer to adaptively select the most task-relevant modality. Simulation results on a real-world multimodal dataset demonstrate the advantages of our proposed model in achieving accurate and efficient beam prediction.
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (21) , 31758-31768
Adaptive and Redundancy-Aware Multimodal Fusion for Sensing-Assisted Beam Prediction
As wireless systems evolve toward higher frequency bands with narrower beams and increased mobility, ensuring accurate beam prediction is crucial for optimizing network performance and improving signal transmission efficiency. Recently, sensing-assisted communication has gained attention for leveraging environmental sensing data to enhance wireless decision-making. However, existing works mainly focus on incorporating more modalities while neglecting the negative impact of modality redundancy. In practice, certain modalities may contribute little to target tasks and introduce unnecessary computational overhead. To address such issue, we will investigate the representational effectiveness of diverse sensing modalities. Driven by numerical observations, we design a novel model architecture, termed as LRCT-MS, that integrates low-rank fusion mechanism with cross-modal transformer to adaptively select the most task-relevant modality. Simulation results on a real-world multimodal dataset demonstrate the advantages of our proposed model in achieving accurate and efficient beam prediction.
- Open Access
- 26 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (17) , 3845
Learning Shared Semantic Representations from WiFi CSI for Unified Multi-Task Human Activity Recognition
Human activity recognition (HAR) plays a critical role in intelligent wireless sensing and mobile edge computing. Compared with traditional vision-based and wearable-based approaches, WiFi channel state information (CSI) enables privacy-preserving and device-free activity perception. CSI encodes human-induced channel variations that serve as a natural basis for activity-oriented semantic analytics over wireless networks. However, existing WiFi CSI-based methods suffer from weak cross-scene generalization, high model complexity, and poor multi-task collaboration. To address these issues, this paper proposes UniSense-CSI, a unified multi-task framework that jointly learns dynamic gesture recognition, static posture classification, and fall detection. Specifically, the proposed framework converts CSI signals into pseudo-RGB images, extracts spatio-temporal features using a customized ConvNeXt backbone, and leverages an improved PerceiverIO module to compress high-dimensional features into a compact latent space. Based on the shared representation, task queries and adapters are introduced to enable parallel multi-task inference within a common architecture. Experiments on public datasets demonstrate that the proposed framework achieves accuracies of 99.64%, 99.66% and 96.09% on the three tasks, respectively, while maintaining low inference latency and favorable edge-deployment capability.
- 26 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (17) , 3845
Open AccessLearning Shared Semantic Representations from WiFi CSI for Unified Multi-Task Human Activity Recognition
Human activity recognition (HAR) plays a critical role in intelligent wireless sensing and mobile edge computing. Compared with traditional vision-based and wearable-based approaches, WiFi channel state information (CSI) enables privacy-preserving and device-free activity perception. CSI encodes human-induced channel variations that serve as a natural basis for activity-oriented semantic analytics over wireless networks. However, existing WiFi CSI-based methods suffer from weak cross-scene generalization, high model complexity, and poor multi-task collaboration. To address these issues, this paper proposes UniSense-CSI, a unified multi-task framework that jointly learns dynamic gesture recognition, static posture classification, and fall detection. Specifically, the proposed framework converts CSI signals into pseudo-RGB images, extracts spatio-temporal features using a customized ConvNeXt backbone, and leverages an improved PerceiverIO module to compress high-dimensional features into a compact latent space. Based on the shared representation, task queries and adapters are introduced to enable parallel multi-task inference within a common architecture. Experiments on public datasets demonstrate that the proposed framework achieves accuracies of 99.64%, 99.66% and 96.09% on the three tasks, respectively, while maintaining low inference latency and favorable edge-deployment capability.
- 27 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 74 (00906778) , 6541-6557
HT-DRIVE: Heterogeneous-Temporal GNN-Aided Multi-Agent Reinforcement Learning for Resource Allocation in Platoon-Based V2X Networks
Platoon-based Vehicle-to-Everything (V2X) is a promising communication paradigm for intelligent vehicular networks. In such networks, each platoon leader (PL) engages in a high-throughput vehicle-to-infrastructure (V2I) link to deliver entertainment data, while maintaining low-latency and high-reliability vehicle-to-vehicle (V2V) links with multiple platoon members (PMs) to convey safety-critical information, thus posing conflicting performance requirements. Besides, the temporal variations and co-channel interferences further complicate the problem. To address these issues, we propose a heterogeneity-aware and dynamics-adaptive resource allocation framework, termed HT-DRIVE, which integrates an advanced heterogeneous temporal graph neural network (HTGNN) with a tailored multi-agent reinforcement learning (MARL) mechanism. Specifically, we design an HTGNN module to capture heterogeneous and temporal patterns within the platoon-based V2X network, generating structured embeddings that serve as state inputs to a modified multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm for downstream decision-making. To facilitate end-to-end optimization, several enhancements are also incorporated, including a hybrid action design to jointly handle discrete subchannel selection and continuous power control, a dual-critics architecture with reduced computational complexity, and customized loss functions for collaborative training. Extensive simulations demonstrate that HT-DRIVE significantly outperforms MATD3, multi-agent deep deterministic policy gradient (MADDPG) and heuristic methods under various settings, validating its effectiveness and robustness.
- 27 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 74 (00906778) , 6541-6557
HT-DRIVE: Heterogeneous-Temporal GNN-Aided Multi-Agent Reinforcement Learning for Resource Allocation in Platoon-Based V2X Networks
Platoon-based Vehicle-to-Everything (V2X) is a promising communication paradigm for intelligent vehicular networks. In such networks, each platoon leader (PL) engages in a high-throughput vehicle-to-infrastructure (V2I) link to deliver entertainment data, while maintaining low-latency and high-reliability vehicle-to-vehicle (V2V) links with multiple platoon members (PMs) to convey safety-critical information, thus posing conflicting performance requirements. Besides, the temporal variations and co-channel interferences further complicate the problem. To address these issues, we propose a heterogeneity-aware and dynamics-adaptive resource allocation framework, termed HT-DRIVE, which integrates an advanced heterogeneous temporal graph neural network (HTGNN) with a tailored multi-agent reinforcement learning (MARL) mechanism. Specifically, we design an HTGNN module to capture heterogeneous and temporal patterns within the platoon-based V2X network, generating structured embeddings that serve as state inputs to a modified multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm for downstream decision-making. To facilitate end-to-end optimization, several enhancements are also incorporated, including a hybrid action design to jointly handle discrete subchannel selection and continuous power control, a dual-critics architecture with reduced computational complexity, and customized loss functions for collaborative training. Extensive simulations demonstrate that HT-DRIVE significantly outperforms MATD3, multi-agent deep deterministic policy gradient (MADDPG) and heuristic methods under various settings, validating its effectiveness and robustness.
- 6 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (24) , 53788-53799
Robust Secure Resource Allocation for MISO-Based SR Systems With HWIs and Channel Uncertainties
Resource allocation (RA) has been considered as a key technique to achieve the optimal system performance in symbiotic radio (SR) systems by optimizing system parameters. However, most of the existing works only consider the ideal hardware conditions or perfect channel information, where the system performance of the above algorithms may be degraded under imperfect channel state information (CSI) (e.g., channel estimation errors) and unideal hardware conditions (e.g., distortion noises). To improve transmission robustness and information security, in this paper, we study the robust secure RA problem for a multiple-input single-output (MISO) SR system under channel uncertainties and hardware impairments (HWIs) with an eavesdropper (Eve). The robust RA problem with bounded channel uncertainties is formulated to maximize the total energy efficiency (EE) of the system under the minimum energy-harvesting constraint of each backscatter device (BD), the maximum transmit power constraint of the primary base station (PBS), the minimum secrecy rate of each BD, the decoding constraint, as well as the reflection coefficient constraint. To address the nonconvex optimization problem, the original robust RA problem with the infinite constraints is converted into a deterministic one via a worst case approach. Then, the objective function is transformed into a nonfractional form using Dinkelbach’s method. After that, the above problem is converted into a convex problem based on $S$ -Procedure and the eigenvalue decomposition approach, and an iterative-based robust secure RA algorithm is proposed via an alternating optimization principle. Simulation results demonstrate that the proposed algorithm has lower outage probabilities and higher EE compared to the nonrobust algorithm and the RA algorithm without HWIs.
- 6 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (24) , 53788-53799
Robust Secure Resource Allocation for MISO-Based SR Systems With HWIs and Channel Uncertainties
Resource allocation (RA) has been considered as a key technique to achieve the optimal system performance in symbiotic radio (SR) systems by optimizing system parameters. However, most of the existing works only consider the ideal hardware conditions or perfect channel information, where the system performance of the above algorithms may be degraded under imperfect channel state information (CSI) (e.g., channel estimation errors) and unideal hardware conditions (e.g., distortion noises). To improve transmission robustness and information security, in this paper, we study the robust secure RA problem for a multiple-input single-output (MISO) SR system under channel uncertainties and hardware impairments (HWIs) with an eavesdropper (Eve). The robust RA problem with bounded channel uncertainties is formulated to maximize the total energy efficiency (EE) of the system under the minimum energy-harvesting constraint of each backscatter device (BD), the maximum transmit power constraint of the primary base station (PBS), the minimum secrecy rate of each BD, the decoding constraint, as well as the reflection coefficient constraint. To address the nonconvex optimization problem, the original robust RA problem with the infinite constraints is converted into a deterministic one via a worst case approach. Then, the objective function is transformed into a nonfractional form using Dinkelbach’s method. After that, the above problem is converted into a convex problem based on $S$ -Procedure and the eigenvalue decomposition approach, and an iterative-based robust secure RA algorithm is proposed via an alternating optimization principle. Simulation results demonstrate that the proposed algorithm has lower outage probabilities and higher EE compared to the nonrobust algorithm and the RA algorithm without HWIs.
- 12 September 2025
- Published by Springer Nature
- p. 323-335
GLFMamba-U: Global-Local Fused Mamba-Unet
- 10 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (6) , 7755-7769
Attentive Radiate Graph for Pedestrian Trajectory Prediction in Disconnected Manifolds
Pedestrian trajectory prediction grapples with the demanding feat of modeling complex interactions and learning multimodal distribution to navigate different human-centric environments. Despite superior performance in reducing distance-based metrics, recent works tend to predict out-of-distribution trajectories, as the distribution of forthcoming paths comprises a blend of various manifolds that may be disconnected. These unrealistic trajectories can potentially jeopardize the safety of traffic participants and result in significant damage. To meet these challenges, we propose DMPred, a graph-based generator adversarial network that generates realistic multimodal trajectory predictions by better modeling the social interactions of pedestrians across different scenes in disconnected manifolds. The core of DMPred is an attentive radiate graph sequence constructed by considering the localized influence radiating from pedestrian movements, which is followed by a spatiotemporal extractor that stores and reuses potentially forgotten neighboring pedestrian information to allow for better extraction of complex interactions. Additionally, a collection of generators is utilized for forecasting, which incorporates spectral clustering on trajectories during the prior learning process of multiple generators to help reduce model redundancy and enhance flexibility for various prediction scenarios. Through extensive experiments on multiple real-world and simulation datasets, we demonstrate that DMPred obtains highly competitive results with efficacy in predicting realistic multimodal trajectories.
- 10 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (6) , 7755-7769
Attentive Radiate Graph for Pedestrian Trajectory Prediction in Disconnected Manifolds
Pedestrian trajectory prediction grapples with the demanding feat of modeling complex interactions and learning multimodal distribution to navigate different human-centric environments. Despite superior performance in reducing distance-based metrics, recent works tend to predict out-of-distribution trajectories, as the distribution of forthcoming paths comprises a blend of various manifolds that may be disconnected. These unrealistic trajectories can potentially jeopardize the safety of traffic participants and result in significant damage. To meet these challenges, we propose DMPred, a graph-based generator adversarial network that generates realistic multimodal trajectory predictions by better modeling the social interactions of pedestrians across different scenes in disconnected manifolds. The core of DMPred is an attentive radiate graph sequence constructed by considering the localized influence radiating from pedestrian movements, which is followed by a spatiotemporal extractor that stores and reuses potentially forgotten neighboring pedestrian information to allow for better extraction of complex interactions. Additionally, a collection of generators is utilized for forecasting, which incorporates spectral clustering on trajectories during the prior learning process of multiple generators to help reduce model redundancy and enhance flexibility for various prediction scenarios. Through extensive experiments on multiple real-world and simulation datasets, we demonstrate that DMPred obtains highly competitive results with efficacy in predicting realistic multimodal trajectories.
- 28 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (7) , 1869-1873
Federated Learning-Aided Dual-Side Channel Estimation in Multi-User Massive MIMO Systems
In massive multiple-input multiple-output (MIMO) systems, attaining precise channel state information (CSI) tends to cause significant pilot overhead and prohibitive computational complexity. In light of this, data-driven approaches have been developed by training neural network models that recover channels from received signals. Since the conventional centralized learning (CL) would bring substantial communication overhead and privacy risks, federated learning (FL) has emerged as a promising alternative, allowing the base station (BS) to collect only model parameters from users, without the whole dataset. However, the previous work focus solely on either downlink or uplink channel estimation, neglecting the potential interplay between such two processes. In this letter, we design a consistent FL framework to estimate both the CSI at the receiver side (CSIR) and the CSI at the transmitter side (CSIT). Specifically, the beamformers at both end during preamble phases are delicately designed, which enables the parameters trained from the downlink models to be fully utilized for uplink channel estimation, thereby further reducing the communication overhead. Simulation results demonstrates that the proposed approach achieves better estimation performance compared to the existing analytical methods, and shows pleasurable robustness adapted to varying environments.
- 28 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (7) , 1869-1873
Federated Learning-Aided Dual-Side Channel Estimation in Multi-User Massive MIMO Systems
In massive multiple-input multiple-output (MIMO) systems, attaining precise channel state information (CSI) tends to cause significant pilot overhead and prohibitive computational complexity. In light of this, data-driven approaches have been developed by training neural network models that recover channels from received signals. Since the conventional centralized learning (CL) would bring substantial communication overhead and privacy risks, federated learning (FL) has emerged as a promising alternative, allowing the base station (BS) to collect only model parameters from users, without the whole dataset. However, the previous work focus solely on either downlink or uplink channel estimation, neglecting the potential interplay between such two processes. In this letter, we design a consistent FL framework to estimate both the CSI at the receiver side (CSIR) and the CSI at the transmitter side (CSIT). Specifically, the beamformers at both end during preamble phases are delicately designed, which enables the parameters trained from the downlink models to be fully utilized for uplink channel estimation, thereby further reducing the communication overhead. Simulation results demonstrates that the proposed approach achieves better estimation performance compared to the existing analytical methods, and shows pleasurable robustness adapted to varying environments.
- Open Access
- 20 November 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 18 (9) , 1-31
A Survey on Knowledge Graph Related Research in Smart City Domain
Knowledge graph employs the specific graph structure to store knowledge in the form of entities, relations, attributes, and so forth, which can effectively represent correlations among data and has been applied in many fields, including search engine optimization, intelligent question answering, and recommendation systems. In this article, we mainly focus on the research and application of the domain-specific knowledge graph in the field of the smart city, which has not been fully paid attention to. Currently, the major problem faced by the smart city lies in data mining and proper application. On the one hand, data are usually stored by government management departments, which creates challenges such as high data storing overhead and inefficient data usage. On the other hand, data cannot be coordinated and collaborated between different city management systems, because data silos exist. By constructing the corresponding knowledge graph, the data of urban traffic, services, and public resources are integrated to provide help for city builders and managers to make important decisions. Therefore, we will review the related literature on the knowledge graph existing in the smart city domain to expore reasearch scopes. Specifically, we will analyze and summarize knowledge graph construction research in the field of smart cities from four perspectives, i.e., smart city ontology, urban data processing, urban knowledge graph construction, and their application. Finally, the research limitations and prospects of the urban knowledge graph are provided.
- 20 November 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 18 (9) , 1-31
Open AccessA Survey on Knowledge Graph Related Research in Smart City Domain
Knowledge graph employs the specific graph structure to store knowledge in the form of entities, relations, attributes, and so forth, which can effectively represent correlations among data and has been applied in many fields, including search engine optimization, intelligent question answering, and recommendation systems. In this article, we mainly focus on the research and application of the domain-specific knowledge graph in the field of the smart city, which has not been fully paid attention to. Currently, the major problem faced by the smart city lies in data mining and proper application. On the one hand, data are usually stored by government management departments, which creates challenges such as high data storing overhead and inefficient data usage. On the other hand, data cannot be coordinated and collaborated between different city management systems, because data silos exist. By constructing the corresponding knowledge graph, the data of urban traffic, services, and public resources are integrated to provide help for city builders and managers to make important decisions. Therefore, we will review the related literature on the knowledge graph existing in the smart city domain to expore reasearch scopes. Specifically, we will analyze and summarize knowledge graph construction research in the field of smart cities from four perspectives, i.e., smart city ontology, urban data processing, urban knowledge graph construction, and their application. Finally, the research limitations and prospects of the urban knowledge graph are provided.
- 1 August 2024
- Published by Springer Nature
- p. 275-286
Attention Enhanced Transformer for Multi-agent Trajectory Prediction
- 1 August 2024
- Published by Springer Nature
- p. 275-286
Attention Enhanced Transformer for Multi-agent Trajectory Prediction