IEEE Transactions on Consumer Electronics最新文献_第10页

Adaptive Federated Learning for Health-Centric Hygiene Monitoring Using LoRa Network 基于LoRa网络的以健康为中心的卫生监测自适应联邦学习

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-03-27 DOI: 10.1109/TCE.2026.3678151

Shubham Pandey;Hari Prabhat Gupta;Rahul Mishra;Priya Gautam;Seela Veerabhadreswara Rao

{"title":"Adaptive Federated Learning for Health-Centric Hygiene Monitoring Using LoRa Network","authors":"Shubham Pandey;Hari Prabhat Gupta;Rahul Mishra;Priya Gautam;Seela Veerabhadreswara Rao","doi":"10.1109/TCE.2026.3678151","DOIUrl":"https://doi.org/10.1109/TCE.2026.3678151","url":null,"abstract":"Hygiene monitoring is essential for timely cleaning and safer operation of healthcare facilities, public washrooms, and similar shared spaces, helping reduce exposure to unhygienic conditions and improve user well-being. However, traditional methods are labor-intensive and lack real-time insights. Existing IoT-based solutions centralize raw data on servers, which raises privacy concerns, increases communication overhead, and limits scalability. This paper presents a health-centric hygiene monitoring system that continuously measures ambient air quality (NH3, H2S, eCO2, TVOC), environmental conditions (temperature and humidity), and motion-based usage patterns in shared washrooms to predict user-reported hygiene ratings reflecting perceived cleanliness. We design an adaptive federated learning framework that keeps raw data local to the device to preserve privacy while considering device heterogeneity and resource constraints. The system uses a LoRa network for communication, enabling long-range, low-power connectivity that supports battery-operated deployments and reduces infrastructure cost. Additionally, the framework optimizes LoRa transmission parameters, groups devices based on computational and link conditions, and applies adaptive model compression with knowledge distillation so that constrained devices can participate using compressed models. A real-world deployment in hostel washrooms demonstrates that the proposed system achieves high classification accuracy for user-reported ratings while significantly reducing communication overhead, indicating its suitability for scalable, privacy-preserving hygiene monitoring in the constrained environments.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"6660-6670"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Blockchain-Assisted Holographic Counterparts for Secure Consumer Electronics in Healthcare 4.0 医疗4.0中安全消费电子产品的区块链辅助全息对应物

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-03-30 DOI: 10.1109/TCE.2026.3679034

Ashish Kumar;Kakali Chatterjee;Ashish Singh;Abhinav Kumar

{"title":"A Blockchain-Assisted Holographic Counterparts for Secure Consumer Electronics in Healthcare 4.0","authors":"Ashish Kumar;Kakali Chatterjee;Ashish Singh;Abhinav Kumar","doi":"10.1109/TCE.2026.3679034","DOIUrl":"https://doi.org/10.1109/TCE.2026.3679034","url":null,"abstract":"Consumer Healthcare Devices (CHD) in Healthcare 4.0 (HC 4.0) increasingly generate continuous physiological data that are transformed into 3-dimensional (3D) holographic visualizations for remote monitoring, diagnosis, and clinical decision support. However, existing Internet of Medical Things (IoMT) and blockchain (BC)-based healthcare systems protect data storage and access but do not verify the integrity, freshness, or provenance of holographic patient representations, leaving such visualizations vulnerable to spoofing, replay, and slice-level tampering. This paper proposes a Blockchain-Assisted Holographic Counterpart (BAHC) framework that cryptographically binds wearable devices to holographic updates using PUF-derived Holographic Authentication Tokens (HAT), enforces slice-level integrity through a Merkle-Hologram-Commitment Tree (Merkle-HC Tree), and anchors updates on a permissioned Proof-of-Authority (PoA) BC. Privacy-preserving access control and verification are achieved using Ciphertext-Policy Attribute-Based Encryption (CP-ABE) and Zero-Knowledge Proofs (ZKPs). The framework is evaluated on a controlled experimental testbed emulating 500 concurrent patient streams using independent public physiological datasets and public MRI volumes for synthetic hologram generation, measuring end-to-end latency, anomaly detection performance, rendering efficiency, and blockchain throughput under up to 100 validators. Experimental results show a 68.6% reduction in holographic rendering latency, a 34% reduction in diagnostic latency, a relative 27% improvement in anomaly detection performance, and sustained throughput close to 500 Transactions Per Second (TPS), demonstrating that BAHC provides a scalable and trustworthy foundation for secure holographic monitoring in HC 4.0 systems.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"6704-6711"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Text-Prompted Cross-Modal Feature Fusion Network for RGB-D Salient Defect Detection 文本提示跨模态特征融合网络用于RGB-D显著缺陷检测

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-02-13 DOI: 10.1109/TCE.2026.3664404

Lisha Cui;Feidie Li;Zhaoyang Ge;Linlin Zhai;Luchao Ma;Xiaoheng Jiang;Mingliang Xu

{"title":"Text-Prompted Cross-Modal Feature Fusion Network for RGB-D Salient Defect Detection","authors":"Lisha Cui;Feidie Li;Zhaoyang Ge;Linlin Zhai;Luchao Ma;Xiaoheng Jiang;Mingliang Xu","doi":"10.1109/TCE.2026.3664404","DOIUrl":"https://doi.org/10.1109/TCE.2026.3664404","url":null,"abstract":"Depth images supplement RGB images with rich object boundaries and shapes. Unfortunately, existing RGB-D based salient defect detection algorithms typically perform directly fusion of cross-modal features, ignoring the impact of depth image quality on model performance. Based on this, we propose a text-prompted cross-modal feature fusion network for RGB-D salient defect detection, abbreviated as TCFNet. To reduce the degradation in detection performance caused by poor-quality depth images, we propose a Text-Prompted Image Quality Assessment module based on the vision-language model CLIP to assess the quality of depth images and dynamically adjust the fusion contribution ratio of depth features. Meanwhile, a Cross-Modal Feature Fusion Module based on the relation-aware and feature-wise attention mechanism is proposed to enhance geometric shapes in depth features and textures and colors in RGB features. Additionally, to deal with discontinuity or fragmentation in the predicted salient maps, we further introduce a Defect Skeleton Guidance Module to provide stable and salient position information for the model, thereby enhancing the coherence of the predicted maps. The extensive experiments on defect detection dataset NEU RSDDS-AUG and general multi-modal dataset ReDWeb-S demonstrate the effectiveness and generalizability of the proposed model.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"5368-5378"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836859","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Tensor-Train-Based Tensor Union for Data Fusion With Its Multi-Modal Accurate Prediction 基于张量训练的数据融合张量联合及其多模态精确预测

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-01-27 DOI: 10.1109/TCE.2026.3657814

Jing Guo;Huazhong Liu;Jihong Ding;Xiaotong Zhou;Weimin Li;Jiawei Wang

{"title":"Tensor-Train-Based Tensor Union for Data Fusion With Its Multi-Modal Accurate Prediction","authors":"Jing Guo;Huazhong Liu;Jihong Ding;Xiaotong Zhou;Weimin Li;Jiawei Wang","doi":"10.1109/TCE.2026.3657814","DOIUrl":"https://doi.org/10.1109/TCE.2026.3657814","url":null,"abstract":"Influenced by the prevalence of mobile edge computing (MEC), massive consumption data is generated from customer applications. Tensor-based data representation and fusion methods are considered to effectively handle these heterogeneous consumption big data. However, the curse of dimensionality caused by high-order tensors and the limited processing power at the edge restrict the operation of data fusion on mobile edge devices. Therefore, this paper focuses on proposing a tensor-train (TT)-based lightweight data fusion method that suitable for applying in the MEC paradigm and diverse scenarios with its multi modal accurate prediction. First, we develop a tensor union method for data fusion in various scenarios, including addition, averaging and Hadamard product. Then, a Union-Then-Decomposition in TT format (UTD-TT) algorithm is presented based on the proposed tensor union. Additionally, we propose three TT-based tensor union approaches appropriately to obtain fusion tensors in lower order space. Furthermore, a Union-By-Decomposition in TT format (UBD-TT) algorithm is presented to achieve TT-based lightweight data fusion. Extensive experimental results indicate that UTD-TT and UBD-TT have better prediction accuracy in predicting user behavior compared to the non-union approach. Besides, UBD-TT outperforms UTD-TT in computational efficiency while maintaining similar prediction accuracies.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7490-7501"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

AMPD: Adaptive Motion Planning Diffusion for Robot Motion Planning in Complex Environments 复杂环境下机器人运动规划的自适应运动规划扩散

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-03-12 DOI: 10.1109/TCE.2026.3673342

Xiangxi Chen;Yuzhi Jiang;Renzhi Lu

{"title":"AMPD: Adaptive Motion Planning Diffusion for Robot Motion Planning in Complex Environments","authors":"Xiangxi Chen;Yuzhi Jiang;Renzhi Lu","doi":"10.1109/TCE.2026.3673342","DOIUrl":"https://doi.org/10.1109/TCE.2026.3673342","url":null,"abstract":"Robot motion planning is a core technology for complex manipulation tasks, widely applied in industrial automation, human-robot collaboration, and service robotics. Traditional planning methods face a dilemma between efficiency and solution quality under complex constraints: sampling-based methods can be inefficient, optimization-based methods may get trapped in local optima, and learning-based methods often generalize poorly. Recently, diffusion-model-based planners have achieved high-quality trajectory generation and can incorporate constraints via gradient guidance. However, most existing approaches apply step-wise guidance at every reverse step, causing excessive inference time; moreover, guidance in early high-noise stages is often low-fidelity and yields limited benefit, wasting computation. We propose AMPD (Adaptive Motion Planning Diffusion), which leverages diffusion noise dynamics to improve real-time efficiency while preserving feasibility. AMPD introduces two complementary designs: 1) a smoothness-aware training objective (SmoothLoss) that shapes the learned prior toward physically plausible trajectories, reducing the need for corrective guidance; and 2) a time-adaptive guidance schedule that allocates guidance computation to denoising stages where gradients are more informative. Experiments on the densely constrained EnvNarrowPassageDense2D benchmark demonstrate that AMPD improves both efficiency and safety over the diffusion baseline MPD, reducing inference time from 1.184s to 0.812s (<inline-formula> <tex-math>$sim 1.46times $ </tex-math></inline-formula> speedup) while increasing the feasible-sample fraction from 73.10% to 91.40%, and decreasing collision intensity from 0.74% to 0.08%. Furthermore, extended evaluations on an extreme maze environment validate the planner’s scalability. AMPD maintains robust performance (89.2% success rate, sub-second inference) where baselines fail due to local optima or timeouts.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7528-7539"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CMA-DF: A Physiological Signal Fusion Method Based on Cross-Modal Alignment and Dynamic Fusion 基于跨模态对齐和动态融合的CMA-DF生理信号融合方法

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-04-09 DOI: 10.1109/TCE.2026.3682466

Xinhui Li;Fuguo Li;Minchao Wu;Hao Chen;Fan Li;Zhao Lv

{"title":"CMA-DF: A Physiological Signal Fusion Method Based on Cross-Modal Alignment and Dynamic Fusion","authors":"Xinhui Li;Fuguo Li;Minchao Wu;Hao Chen;Fan Li;Zhao Lv","doi":"10.1109/TCE.2026.3682466","DOIUrl":"https://doi.org/10.1109/TCE.2026.3682466","url":null,"abstract":"Multimodal emotion recognition based on complementary physiological signals, such as an electroencephalogram (EEG) and eye movement, can effectively reflect human emotional states, showing great potential in fields like rehabilitation monitoring and driving safety management. However, existing methods usually directly map multimodal features into a unified feature space for information fusion, which not only fails to eliminate the heterogeneity between different modalities but also ignores the impact of the modal contribution on the fusion results. To address these limitations, this paper proposes a cross-modal alignment and dynamic fusion (CMA-DF) learning framework to improve the performance of multimodal emotion recognition. Specifically, we first adopt a two-branch feature extractor to respectively extract robust feature representations of EEG signals and eye movement signals. Secondly, a cross-modal alignment network is designed, which reduces distribution differences through collaborative modeling of intermodal interaction relationships and alignment strategies, thus providing a consistent basis for subsequent fusion. Subsequently, a dynamic fusion strategy is introduced to efficiently integrate complementary information from multiple modalities while adaptively balancing the contribution weights of each modality, thereby enhancing the representational capability of the fused features. Finally, experimental results in the SEED and SEED-V datasets show that the proposed method achieves recognition accuracies of 96.69% and 88.03%, respectively. Comparative experiments fully verify its superiority in multimodal feature alignment, fusion, and emotion recognition performance.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7707-7720"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Bridging Static Images and Dynamic Brain Signal for Visual Decoding 桥接静态图像和动态脑信号的视觉解码

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-04-28 DOI: 10.1109/TCE.2026.3688178

Dongjun Liu;Weichen Dai;Honggang Liu;Hangjie Yi;Wanzeng Kong

{"title":"Bridging Static Images and Dynamic Brain Signal for Visual Decoding","authors":"Dongjun Liu;Weichen Dai;Honggang Liu;Hangjie Yi;Wanzeng Kong","doi":"10.1109/TCE.2026.3688178","DOIUrl":"https://doi.org/10.1109/TCE.2026.3688178","url":null,"abstract":"Object perception and recognition capabilities are fundamental to advancing user experience in consumer electronics. Current research decoding visual information from brain signals is a fundamental challenge for understanding human cognition and advancing next-generation consumer-grade brain-machine interfaces (BMIs). Previous brain-vision alignment methods tend to overlook the commonly existing issue of information asymmetry. This limitation arises from the fundamental structural mismatch between static image features and dynamic neural responses, as images represent instantaneous visual content while the brain processes stimuli through continuous temporal, spatial, and spectral dynamics. To address this gap, we propose a simple yet effective multi-level alignment framework that integrates visual biological priors and cognitive dynamic priors. Specifically, we generate Gaussian-blurred image variants to simulate foveated human vision and decompose brain signals according to time-frequency with adaptive attention to enhance visual relevance. These cross-modal variants are then aligned via contrastive learning to achieve reasonable and improved brain-image correspondence. Our method achieved an accuracy rate of 62.5% for the top 1 and 90.7% for the top 5 in the zero-shot brain-to-image retrieval task, surpassing the previous state-of-the-art method by 11.6% and 11.0%, respectively. The results highlight that incorporating visual and cognitive priors into alignment not only improves the interpretability but also enhances decoding accuracy.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7721-7734"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836965","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CineSteady: A Real-Time Hybrid Video Stabilization and Distortion Correction Method CineSteady：一种实时混合视频稳定和失真校正方法

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-03-02 DOI: 10.1109/TCE.2026.3669217

Yan Chen;Zijing Mao;Chang Liu;Jian Tang;Yiduo Zhang;Chenhui Yin;Heng Sun

{"title":"CineSteady: A Real-Time Hybrid Video Stabilization and Distortion Correction Method","authors":"Yan Chen;Zijing Mao;Chang Liu;Jian Tang;Yiduo Zhang;Chenhui Yin;Heng Sun","doi":"10.1109/TCE.2026.3669217","DOIUrl":"https://doi.org/10.1109/TCE.2026.3669217","url":null,"abstract":"The rapid development of the consumer electronics industry has highlighted the significance of video stabilization technologies for devices like smartphones. Nevertheless, State-of-the-Art (SOTA) solutions face significant challenges in real-world deployment, primarily due to limitations in stability, image distortion control, black border elimination, and real-time processing capabilities. To address these challenges, we propose <bold>CineSteady, a novel and comprehensive video stabilization algorithm framework. First, by using gyroscope data and image sequences, CineSteady achieves high-order smoothing of video sequences under complex motion scenarios, delivering a top-tier stabilization experience for users. In addition, we introduce a forward-backward filtering algorithm that rigorously enforces boundary constraints. This approach guarantees black-border-free output within a fixed Field of View (FOV) while maximizing video stability. Moreover, to address practical deployment challenges on mobile devices, we present Optical Lens-based Rolling Shutter Correction (OLRC), a comprehensive camera and distortion correction model. By leveraging physical imaging models and 3D motion estimation, OLRC effectively eliminates the jello effect. Finally, the proposed algorithm demonstrates exceptional computational efficiency, achieving real-time performance on the Xiaomi 15 Ultra powered by the Qualcomm Snapdragon 8 Elite platform, supporting 4K resolution at 60 frames per second. Evaluations in diverse real-world scenarios confirm that CineSteady surpasses SOTA video stabilization methods in user experience, demonstrating significant advantages of the proposed solution.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"5429-5443"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837060","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Safe Multi-Agent Deep Reinforcement Learning-Based Energy Management for Multi-Microgrid Energy Communities 基于安全多智能体深度强化学习的多微电网能源社区能源管理

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-03-31 DOI: 10.1109/TCE.2026.3679622

Mengfan Zhang;Zhipeng Li;Hong Li;Qianwen Xu

{"title":"Safe Multi-Agent Deep Reinforcement Learning-Based Energy Management for Multi-Microgrid Energy Communities","authors":"Mengfan Zhang;Zhipeng Li;Hong Li;Qianwen Xu","doi":"10.1109/TCE.2026.3679622","DOIUrl":"https://doi.org/10.1109/TCE.2026.3679622","url":null,"abstract":"Multi-microgrid (MMG) systems enhance efficient integration of renewables and emerging loads into modern sustainable communities. By facilitating energy sharing among MGs, MMG systems can enhance reliability, improve efficiency, and promote economic operation. In this paper, a projection based safe multi-agent deep reinforcement learning (DRL) algorithm is developed for the decentralized energy management of multi-microgrid energy communities. The safe projection layer integrated into DRL training effectively restricts the agent’s actions to the feasible space, thereby preventing any violations of physical constraints during system operation. The decentralized application of the proposed safe DRL methods achieves the optimal energy management without high-fidelity communication, thereby facilitating practical deployment. The modified MMG test systems are adopted to test the proposed method, and the testing results could verify the effectiveness of the proposed energy management system in achieving cost-effective operation while guaranteeing 100% safety and eliminating the dependence on high-fidelity communication.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"6264-6275"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11458803","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Asymmetric Hierarchical Federated Learning for Vehicular Edge Intelligence: A Knowledge-Gradient Dual Flow Approach 面向车辆边缘智能的非对称分层联邦学习：一种知识梯度双流方法

IF 9.9 2区计算机科学

IEEE Transactions on Consumer Electronics Pub Date : 2026-08-01 Epub Date: 2026-04-07 DOI: 10.1109/TCE.2026.3681623

Zhe Li;Shaopeng Guan

{"title":"Asymmetric Hierarchical Federated Learning for Vehicular Edge Intelligence: A Knowledge-Gradient Dual Flow Approach","authors":"Zhe Li;Shaopeng Guan","doi":"10.1109/TCE.2026.3681623","DOIUrl":"https://doi.org/10.1109/TCE.2026.3681623","url":null,"abstract":"High mobility and non-independent and identically distributed (non-IID) data in the Internet of Vehicles (IoV) severely limit the communication efficiency and convergence of federated learning. This paper proposes an adaptive hierarchical federated learning framework that jointly improves communication efficiency, model accuracy, and fairness through downlink knowledge guidance and uplink gradient compression. The framework forms an asymmetric training loop in which multi-dimensional knowledge distillation transfers edge logits and intermediate features to guide heterogeneous vehicles toward fast feature alignment, thereby mitigating non-IID conflicts. On this basis, an adaptive gradient compression mechanism enables roadside units to measure the cosine similarity and magnitude of local updates in real time and to compute sparsification thresholds dynamically so as to reduce communication cost while preserving model accuracy. A dynamic client selection strategy that combines long-term reputation with <inline-formula> <tex-math>$epsilon $ </tex-math></inline-formula>-greedy exploration further decouples efficiency from fairness and alleviates the winner-takes-all problem. Experiments based on real traffic trajectories and vision benchmarks show that the proposed framework reduces communication overhead by 94.3% compared with standard federated averaging. It achieves 98.17% test accuracy on GTSRB and improves accuracy by up to 6.01 percentage points over the best baseline on CIFAR-100, while effectively exploiting data heterogeneity for additional performance gains.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"5650-5660"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837116","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0