Pub Date : 2025-08-19DOI: 10.1109/OJVT.2025.3600512
Qiang Wang;Yongchong Xue;Shuchang Lyu;Guangliang Cheng;Shaoyan Yang;Xin Jin
Effective autonomous driving systems require a delicate balance of high precision, efficient design, and immediate response capabilities. This study presents MTFENet, a cutting-edge multi-task deep learning model that optimizes network architecture to harmonize speed and accuracy for critical tasks such as object detection, drivable area segmentation, and lane line segmentation. Our end-to-end, streamlined multi-task model incorporates an Adaptive Feature Fusion Module (AF$^{2}$M) to manage the diverse feature demands of different tasks. We also introduced a fusion transform module (FTM) to strengthen global feature extraction and a novel detection head to address target loss and confusion. To enhance computational efficiency, we refined the segmentation head design. Experiments on the BDD100k dataset reveal that MTFENet delivers exceptional performance, achieving an mAP50 of 81.5% in object detection, an mIoU of 93.8% in drivable area segmentation, and an IoU of 33.7% in lane line segmentation. Real-world scenario evaluations demonstrate that MTFENet substantially outperforms current state-of-the-art models across multiple tasks, highlighting its superior adaptability and swift response. These results underscore that MTFENet not only leads in precision and speed but also bolsters the reliability and adaptability of autonomous driving systems in navigating complex road conditions.
{"title":"MTFENet: A Multi-Task Autonomous Driving Network for Real-Time Target Perception","authors":"Qiang Wang;Yongchong Xue;Shuchang Lyu;Guangliang Cheng;Shaoyan Yang;Xin Jin","doi":"10.1109/OJVT.2025.3600512","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3600512","url":null,"abstract":"Effective autonomous driving systems require a delicate balance of high precision, efficient design, and immediate response capabilities. This study presents MTFENet, a cutting-edge multi-task deep learning model that optimizes network architecture to harmonize speed and accuracy for critical tasks such as object detection, drivable area segmentation, and lane line segmentation. Our end-to-end, streamlined multi-task model incorporates an Adaptive Feature Fusion Module (AF<inline-formula><tex-math>$^{2}$</tex-math></inline-formula>M) to manage the diverse feature demands of different tasks. We also introduced a fusion transform module (FTM) to strengthen global feature extraction and a novel detection head to address target loss and confusion. To enhance computational efficiency, we refined the segmentation head design. Experiments on the BDD100k dataset reveal that MTFENet delivers exceptional performance, achieving an mAP50 of 81.5% in object detection, an mIoU of 93.8% in drivable area segmentation, and an IoU of 33.7% in lane line segmentation. Real-world scenario evaluations demonstrate that MTFENet substantially outperforms current state-of-the-art models across multiple tasks, highlighting its superior adaptability and swift response. These results underscore that MTFENet not only leads in precision and speed but also bolsters the reliability and adaptability of autonomous driving systems in navigating complex road conditions.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2406-2423"},"PeriodicalIF":4.8,"publicationDate":"2025-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11130405","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145027938","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-15DOI: 10.1109/OJVT.2025.3599570
José Rodríguez-Piñeiro;Zhongxiang Wei;Jingjing Wang;Carlos A. Gutiérrez;Luis M. Correia
With the developments on Advanced Driver-Assistance Systems (ADAS), autonomous driving and purely unmanned vehicles, such as the Unmanned Aerial Vehicles (UAVs), the pressure on the requirements for Vehicle-to-Everything (V2X) communications has drastically increased during the last few years. However, a new and appealing horizon is open for V2X applications with the advent of the Sixth Generation (6G) of mobile communications. In this paper, we present a review of V2X standards to offer a holistic perspective on the evolution of this technology toward 6G. Then, the key technological enablers for the 6G V2X are identified, namely (a) Non-Terrestrial Networks (NTNs), (b) Ultra Reliable Low Latency Communications (URLLC), (c) Artificial Intelligence (AI), (d) Integrated Sensing And Communications (ISAC) and (e) propagation channel modeling. For each of these technological enablers, the most recent proposals are thoroughly studied and the open challenges and opportunities identified. Our paper intends to serve as a timely roadmap for the development of future 6G V2X networks.
{"title":"6G-Enabled Vehicle-to-Everything Communications: Current Research Trends and Open Challenges","authors":"José Rodríguez-Piñeiro;Zhongxiang Wei;Jingjing Wang;Carlos A. Gutiérrez;Luis M. Correia","doi":"10.1109/OJVT.2025.3599570","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3599570","url":null,"abstract":"With the developments on Advanced Driver-Assistance Systems (ADAS), autonomous driving and purely unmanned vehicles, such as the Unmanned Aerial Vehicles (UAVs), the pressure on the requirements for Vehicle-to-Everything (V2X) communications has drastically increased during the last few years. However, a new and appealing horizon is open for V2X applications with the advent of the Sixth Generation (6G) of mobile communications. In this paper, we present a review of V2X standards to offer a holistic perspective on the evolution of this technology toward 6G. Then, the key technological enablers for the 6G V2X are identified, namely <italic>(a)</i> Non-Terrestrial Networks (NTNs), <italic>(b)</i> Ultra Reliable Low Latency Communications (URLLC), <italic>(c)</i> Artificial Intelligence (AI), <italic>(d)</i> Integrated Sensing And Communications (ISAC) and <italic>(e)</i> propagation channel modeling. For each of these technological enablers, the most recent proposals are thoroughly studied and the open challenges and opportunities identified. Our paper intends to serve as a timely roadmap for the development of future 6G V2X networks.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2358-2391"},"PeriodicalIF":4.8,"publicationDate":"2025-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11126933","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145011309","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-11DOI: 10.1109/OJVT.2025.3597730
Noud B. Kanters;Andrés Alayón Glazunov
This paper investigates the application of recently proposed practical subarray (SA)-based hybrid beamforming (HBF) architectures—implemented entirely with passive beamforming networks and switches—for millimeter wave (mmWave) multi-user (MU)-MIMO base stations. Building on this practical hardware platform, we propose a joint SA configuration and signal processing framework that exploits the natural non-uniformity of user locations in 3-D space via elevation domain subsectorization. Specifically, we adapt established channel estimation and HBF techniques to the constraints of switch-based SAs, and introduce a novel 2-stage channel estimator that leverages the unique properties of mmWave channels. System-level simulations in realistic line-of-sight (LoS) and non-line-of-sight (NLoS) propagation scenarios demonstrate that the proposed solution delivers strong performance with low complexity, providing a viable path toward practical, scalable mmWave MU-MIMO deployments. In LoS scenarios, using directions-of-arrival-based channel estimation, the proposed framework achieves up to 92.6% of the average spectral efficiency (SE) of a full-digital array antenna with the same number of elements but 4 times more radio frequency chains. In NLoS environments, using the novel 2-stage estimator, this increases up to 99.7%.
{"title":"Steerable Subarrays for Practical mmWave Massive MIMO: Algorithm Design and System-Level Analysis","authors":"Noud B. Kanters;Andrés Alayón Glazunov","doi":"10.1109/OJVT.2025.3597730","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3597730","url":null,"abstract":"This paper investigates the application of recently proposed practical subarray (SA)-based hybrid beamforming (HBF) architectures—implemented entirely with passive beamforming networks and switches—for millimeter wave (mmWave) multi-user (MU)-MIMO base stations. Building on this practical hardware platform, we propose a joint SA configuration and signal processing framework that exploits the natural non-uniformity of user locations in 3-D space via elevation domain subsectorization. Specifically, we adapt established channel estimation and HBF techniques to the constraints of switch-based SAs, and introduce a novel 2-stage channel estimator that leverages the unique properties of mmWave channels. System-level simulations in realistic line-of-sight (LoS) and non-line-of-sight (NLoS) propagation scenarios demonstrate that the proposed solution delivers strong performance with low complexity, providing a viable path toward practical, scalable mmWave MU-MIMO deployments. In LoS scenarios, using directions-of-arrival-based channel estimation, the proposed framework achieves up to 92.6% of the average spectral efficiency (SE) of a full-digital array antenna with the same number of elements but 4 times more radio frequency chains. In NLoS environments, using the novel 2-stage estimator, this increases up to 99.7%.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2224-2235"},"PeriodicalIF":4.8,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11122600","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144918318","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-11DOI: 10.1109/OJVT.2025.3598154
Mohamed El-Emary;Diala Naboulsi;Razvan Stanica
Unmanned aerial vehicle (UAV)-assisted Mobile Edge Computing (MEC) presents a critical trade-off between minimizing user equipment (UE) energy consumption and ensuring high task execution reliability, especially for mission-critical applications.While many frameworks focus on either energy efficiency or resiliency, few address both objectives simultaneously with a structured redundancy model. To bridge this gap, this paper proposes a novel reinforcement learning (RL)-based framework that intelligently distributes computational tasks among UAVs and base stations (BSs). We introduce an $(h+1)$-server permutation strategy that redundantly assigns tasks to multiple edge servers, guaranteeing execution continuity even under partial system failures. An RL agent optimizes the offloading process by leveraging network state information to balance energy consumption with system robustness. Extensive simulations demonstrate the superiority of our approach over state-of-the-art benchmarks. Notably, our proposed framework sustains average UE energy levels above 75% under high user densities, exceeds 95% efficiency with more base stations, and maintains over 90% energy retention when 20 or more UAVs are deployed. Even under high computational loads, it preserves more than 50% of UE energy, outperforming all benchmarks by a significant margin—especially for mid-range task sizes where it leads by over 15–20% in energy efficiency. These findings highlight the potential of our framework to support energy-efficient and failure-resilient services for next-generation wireless networks.
{"title":"Energy Efficient and Resilient Task Offloading in UAV-Assisted MEC Systems","authors":"Mohamed El-Emary;Diala Naboulsi;Razvan Stanica","doi":"10.1109/OJVT.2025.3598154","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3598154","url":null,"abstract":"Unmanned aerial vehicle (UAV)-assisted Mobile Edge Computing (MEC) presents a critical trade-off between minimizing user equipment (UE) energy consumption and ensuring high task execution reliability, especially for mission-critical applications.While many frameworks focus on either energy efficiency or resiliency, few address both objectives simultaneously with a structured redundancy model. To bridge this gap, this paper proposes a novel reinforcement learning (RL)-based framework that intelligently distributes computational tasks among UAVs and base stations (BSs). We introduce an <inline-formula><tex-math>$(h+1)$</tex-math></inline-formula>-server permutation strategy that redundantly assigns tasks to multiple edge servers, guaranteeing execution continuity even under partial system failures. An RL agent optimizes the offloading process by leveraging network state information to balance energy consumption with system robustness. Extensive simulations demonstrate the superiority of our approach over state-of-the-art benchmarks. Notably, our proposed framework sustains average UE energy levels above 75% under high user densities, exceeds 95% efficiency with more base stations, and maintains over 90% energy retention when 20 or more UAVs are deployed. Even under high computational loads, it preserves more than 50% of UE energy, outperforming all benchmarks by a significant margin—especially for mid-range task sizes where it leads by over 15–20% in energy efficiency. These findings highlight the potential of our framework to support energy-efficient and failure-resilient services for next-generation wireless networks.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2236-2254"},"PeriodicalIF":4.8,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11122595","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144918204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-11DOI: 10.1109/OJVT.2025.3597659
Fan Yu;Mingqi Guo;Qi Wang;Pengqi Zhu;Yixiao Tong;José Rodríguez-Piñeiro;Xuefeng Yin
Vehicle-to-vehicle (V2V) wireless communication is vital for intelligent transportation systems (ITSs). The high mobility of transceivers, along with the complex 3D propagation caused by low antenna heights and short communication ranges, present challenges to propagation modeling. Accurate V2V channel models are crucial for capturing these characteristics to design reliable V2V systems. Existing cluster-based V2V channel models neglect Doppler frequency variations in cluster classification, reducing classification and model accuracy. They describe clusters in single snapshot, missing temporal channel stationarity, and their complex structures slow model generation, hampering ITS applications. This paper presents a cluster-based V2V channel model incorporating quasi-stationary segmentation. First, SAGE algorithm extracts Multipath components (MPCs), followed by clustering and tracking. By analyzing clusters' Doppler frequency variations alongside angle, delay, and power changes, clusters are more accurately classified into global, static and dynamic types. Next, the model uses Correlation matrix distances (CMDs) to perform quasi-stationary segments for each cluster type, characterizing their distributions within each segment via inter- and intra-cluster parameters. This simplifies the model structure compared to single-snapshot models, improving generation efficiency. Segment duration and quantity statistics characterize channel stationarity. The model is validated by comparing simulated second-order channel statistics with comparable models and measured data. Its complexity is evaluated by comparing model generation time with alternative models in the literature.
{"title":"A Cluster-Based Channel Model Incorporating Quasi-Stationary Segmentation for Vehicle-to-Vehicle Communications","authors":"Fan Yu;Mingqi Guo;Qi Wang;Pengqi Zhu;Yixiao Tong;José Rodríguez-Piñeiro;Xuefeng Yin","doi":"10.1109/OJVT.2025.3597659","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3597659","url":null,"abstract":"Vehicle-to-vehicle (V2V) wireless communication is vital for intelligent transportation systems (ITSs). The high mobility of transceivers, along with the complex 3D propagation caused by low antenna heights and short communication ranges, present challenges to propagation modeling. Accurate V2V channel models are crucial for capturing these characteristics to design reliable V2V systems. Existing cluster-based V2V channel models neglect Doppler frequency variations in cluster classification, reducing classification and model accuracy. They describe clusters in single snapshot, missing temporal channel stationarity, and their complex structures slow model generation, hampering ITS applications. This paper presents a cluster-based V2V channel model incorporating quasi-stationary segmentation. First, SAGE algorithm extracts Multipath components (MPCs), followed by clustering and tracking. By analyzing clusters' Doppler frequency variations alongside angle, delay, and power changes, clusters are more accurately classified into global, static and dynamic types. Next, the model uses Correlation matrix distances (CMDs) to perform quasi-stationary segments for each cluster type, characterizing their distributions within each segment via inter- and intra-cluster parameters. This simplifies the model structure compared to single-snapshot models, improving generation efficiency. Segment duration and quantity statistics characterize channel stationarity. The model is validated by comparing simulated second-order channel statistics with comparable models and measured data. Its complexity is evaluated by comparing model generation time with alternative models in the literature.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2424-2440"},"PeriodicalIF":4.8,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11122388","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145027987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the advancement of electric vehicles towards intelligence and integration, four-wheel independent drive (FWID) vehicles, characterized by high controllability and structural flexibility, have gained widespread attention. Due to multi-degree-of-freedom coupling characteristics, the FWID constitutes complex nonlinear system, necessitating an adaptive control framework to enhance stability and safety. In this paper, a dual-level hierarchical functional control (DHFC) is proposed for FWID, aiming to exploit the potential of the FWID in achieving coordinated optimization of driving safety and stability. The high-level controller is designed to accurately determine the global stability status of FWID by enhancing both parameter estimation accuracy and safety constraints. A reinforcement learning-enhanced high-order cubature Kalman filter (RL-HCKF) improves adaptability and responsiveness in FWID state estimation. Additionally, a hybrid offline-online region of attraction (ROA) identification mechanism is established to delineate safety constraint boundaries for FWID. Meanwhile, the low-level controller adopts stochastic model predictive control (SMPC) to synthesize wheel-level torque vectoring, with dynamically adjusted constraints to enhance the robustness and safety of FWID under uncertain conditions. Simulation evaluations and hardware-in-the-loop (HIL) tests confirm the effectiveness of the proposed strategy. The results demonstrate that, compared to representative existing methods, the DHFC exhibits superior control stability and disturbance adaptability under various driving conditions.
{"title":"A Dual-Level Hierarchical Functional Control Strategy for Four-Wheel Independent Drive Vehicles: Coordination for Enhanced Stability and Safety","authors":"Zhiqi Guo;Liang Chu;Xiaoxu Wang;Yuhang Xiao;Zixu Wang;Zhuoran Hou","doi":"10.1109/OJVT.2025.3596560","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3596560","url":null,"abstract":"With the advancement of electric vehicles towards intelligence and integration, four-wheel independent drive (FWID) vehicles, characterized by high controllability and structural flexibility, have gained widespread attention. Due to multi-degree-of-freedom coupling characteristics, the FWID constitutes complex nonlinear system, necessitating an adaptive control framework to enhance stability and safety. In this paper, a dual-level hierarchical functional control (DHFC) is proposed for FWID, aiming to exploit the potential of the FWID in achieving coordinated optimization of driving safety and stability. The high-level controller is designed to accurately determine the global stability status of FWID by enhancing both parameter estimation accuracy and safety constraints. A reinforcement learning-enhanced high-order cubature Kalman filter (RL-HCKF) improves adaptability and responsiveness in FWID state estimation. Additionally, a hybrid offline-online region of attraction (ROA) identification mechanism is established to delineate safety constraint boundaries for FWID. Meanwhile, the low-level controller adopts stochastic model predictive control (SMPC) to synthesize wheel-level torque vectoring, with dynamically adjusted constraints to enhance the robustness and safety of FWID under uncertain conditions. Simulation evaluations and hardware-in-the-loop (HIL) tests confirm the effectiveness of the proposed strategy. The results demonstrate that, compared to representative existing methods, the DHFC exhibits superior control stability and disturbance adaptability under various driving conditions.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2255-2271"},"PeriodicalIF":4.8,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11119462","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144934368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-06DOI: 10.1109/OJVT.2025.3596251
Dawei Wang;Hongyan Wang;Weichao Yang;Yixin He;Yi Jin;Li Li;Hongbo Zhao;Xiaoyang Li
Autonomous aerial vehicles (AAVs) can effectively eliminate communication blind zones and establish line-of-sight links with ground vehicles by leveraging their flexible deployment capabilities. Motivated by the above, this paper employs an AAV as a mobile edge computing (MEC) server to provide task offloading services, based on which the non-orthogonal multiple access (NOMA) technology is used in AAV-enabled Internet of Vehicles (IoV). To reduce the MEC offloading delay, we propose a NOMA-enhanced MEC framework for AAV-enabled IoV. More explicitly, we formulate a total offloading delay minimization problem by optimizing the transmit power and the AAV position. To tackle the non-convex problem, we decouple it into two sub-problems: power allocation and AAV position optimization. Specifically, the power allocation is optimized via the successive convex optimization algorithm, and the AAV position is adjusted using the improved particle swarm optimization-genetic algorithm (PSO-GA). Then, we propose an iterative optimization algorithm to alternately iterate these two processes to find the optimal solution. Next, we analyze the achievable offloading probability of the NOMA-MEC scheme compared with the OMA-MEC scheme and derive its asymptotic expressions under high signal-to-noise ratio (SNR) conditions. Finally, simulation results indicate that the proposed scheme outperforms existing methods in reducing total offloading delay while validating the accuracy of performance analysis.
{"title":"Mobile Edge Computing for AAV-Enabled Internet of Vehicles With NOMA: Delay Optimization and Performance Analysis","authors":"Dawei Wang;Hongyan Wang;Weichao Yang;Yixin He;Yi Jin;Li Li;Hongbo Zhao;Xiaoyang Li","doi":"10.1109/OJVT.2025.3596251","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3596251","url":null,"abstract":"Autonomous aerial vehicles (AAVs) can effectively eliminate communication blind zones and establish line-of-sight links with ground vehicles by leveraging their flexible deployment capabilities. Motivated by the above, this paper employs an AAV as a mobile edge computing (MEC) server to provide task offloading services, based on which the non-orthogonal multiple access (NOMA) technology is used in AAV-enabled Internet of Vehicles (IoV). To reduce the MEC offloading delay, we propose a NOMA-enhanced MEC framework for AAV-enabled IoV. More explicitly, we formulate a total offloading delay minimization problem by optimizing the transmit power and the AAV position. To tackle the non-convex problem, we decouple it into two sub-problems: power allocation and AAV position optimization. Specifically, the power allocation is optimized via the successive convex optimization algorithm, and the AAV position is adjusted using the improved particle swarm optimization-genetic algorithm (PSO-GA). Then, we propose an iterative optimization algorithm to alternately iterate these two processes to find the optimal solution. Next, we analyze the achievable offloading probability of the NOMA-MEC scheme compared with the OMA-MEC scheme and derive its asymptotic expressions under high signal-to-noise ratio (SNR) conditions. Finally, simulation results indicate that the proposed scheme outperforms existing methods in reducing total offloading delay while validating the accuracy of performance analysis.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2317-2331"},"PeriodicalIF":4.8,"publicationDate":"2025-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11119075","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145011312","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-04DOI: 10.1109/OJVT.2025.3595200
Yang Zhao;SI-YU Zhang;Yuexia Zhang;Gongpu Wang;Behnam Shahrrava
Orthogonal Frequency Division Multiplexing with Index Modulation (OFDM-IM) is regarded as a promising candidate for next generation communications due to its remarkable efficiency and flexibility. In the field of wireless communications, deep learning, particularly Convolutional Neural Networks (CNNs), has been extensively utilized for tasks such as channel estimation and signal detection. However, CNNs' limited receptive field growth poses a challenge in capturing long range dependencies. To achieve efficient deep learning based OFDM-IM detection, this paper proposes two novel OFDM-IM signal detection networks that integrate wavelet transforms with CNNs (WTConv). The first proposed network, referred to as Dual Stage Wavelet Convolutions (DS-WTConv), adopts a dual stage architecture. It comprises an Index Feature Extraction Sub-Network (IdxNet) and a Signal Feature Reconstruction Sub-Network (DetNet). The second network, named Single Network Wavelet Convolutions (SN-WTConv), features a more compact single stage design that combines wavelet convolution and CNN layers. Extensive simulation results demonstrate that both the DS-WTConv and SN-WTConv networks exhibit superior bit error rate (BER) performance and lower computational complexity compared to existing conventional and deep learning-based approaches. Compared to the existing deep learning based detection schemes, the proposed WTConv-based networks reduce the BER by up to 35.3%, and the running time by up to 30.1%. Compared to the optimal Maximum likelihood (ML) method, the proposed DS-WTConv and SN-WTConv achieve approximately 19.2 times and 11.3 times faster runtime, respectively.
{"title":"Novel Wavelet Convolutional Neural Networks for Signal Detection in OFDM-IM Systems","authors":"Yang Zhao;SI-YU Zhang;Yuexia Zhang;Gongpu Wang;Behnam Shahrrava","doi":"10.1109/OJVT.2025.3595200","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3595200","url":null,"abstract":"Orthogonal Frequency Division Multiplexing with Index Modulation (OFDM-IM) is regarded as a promising candidate for next generation communications due to its remarkable efficiency and flexibility. In the field of wireless communications, deep learning, particularly Convolutional Neural Networks (CNNs), has been extensively utilized for tasks such as channel estimation and signal detection. However, CNNs' limited receptive field growth poses a challenge in capturing long range dependencies. To achieve efficient deep learning based OFDM-IM detection, this paper proposes two novel OFDM-IM signal detection networks that integrate wavelet transforms with CNNs (WTConv). The first proposed network, referred to as Dual Stage Wavelet Convolutions (DS-WTConv), adopts a dual stage architecture. It comprises an Index Feature Extraction Sub-Network (IdxNet) and a Signal Feature Reconstruction Sub-Network (DetNet). The second network, named Single Network Wavelet Convolutions (SN-WTConv), features a more compact single stage design that combines wavelet convolution and CNN layers. Extensive simulation results demonstrate that both the DS-WTConv and SN-WTConv networks exhibit superior bit error rate (BER) performance and lower computational complexity compared to existing conventional and deep learning-based approaches. Compared to the existing deep learning based detection schemes, the proposed WTConv-based networks reduce the BER by up to 35.3%, and the running time by up to 30.1%. Compared to the optimal Maximum likelihood (ML) method, the proposed DS-WTConv and SN-WTConv achieve approximately 19.2 times and 11.3 times faster runtime, respectively.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2210-2223"},"PeriodicalIF":4.8,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11107403","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144896781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-31DOI: 10.1109/OJVT.2025.3593944
Si-Yu Zhang;Jia-Qi Zhang;Xin-Wei Yue;Chao-Wei Wang
Short to medium length polar codes achieve inferior decoding performance than other advanced channel codes under successive cancellation (SC). Sophisticated polar decoding enhances the corresponding performance while degrading the coding rate and complexity. For better decoding performance and efficiency, this paper presents a polar coding scheme with selected index modulation (PC-SIM). At the transmitter, PC-SIM integrates the concept of index modulation (IM) into polar encoding, using the indices of inactive unfrozen positions (IUPs) to carry implicit information. To boost coding rate without sacrificing decoding performance, PC-SIM selects more reliable unfrozen positions for IM and adds inactive information bits (IIBs) to offset rate losses. Additionally, Walsh-Hadamard Transform (WHT) is incorporated to lower the high peak-to-average power ratio (PAPR) in multi-carrier systems and reduce interference. At the receiver, PC-SIM performs polar decoding followed by repetition decoding to obtain index bits and information bits. Simulation results indicate that in Orthogonal Frequency Division Multiplexing (OFDM) systems, compared to conventional polar codes and existing IM-aided polar coding schemes, the proposed PC-SIM scheme significantly improves error performance, coding rate, and PAPR reduction. The proposed PC-SIM achieves around 0.3 dB over the conventional CRC-aided polar codes and IM-aided polar codes with higher coding rate at the bit error ratio (BER) of $4times 10^{-4}$.
{"title":"A Polar Coding Scheme With Selected Index Modulation","authors":"Si-Yu Zhang;Jia-Qi Zhang;Xin-Wei Yue;Chao-Wei Wang","doi":"10.1109/OJVT.2025.3593944","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3593944","url":null,"abstract":"Short to medium length polar codes achieve inferior decoding performance than other advanced channel codes under successive cancellation (SC). Sophisticated polar decoding enhances the corresponding performance while degrading the coding rate and complexity. For better decoding performance and efficiency, this paper presents a polar coding scheme with selected index modulation (PC-SIM). At the transmitter, PC-SIM integrates the concept of index modulation (IM) into polar encoding, using the indices of inactive unfrozen positions (IUPs) to carry implicit information. To boost coding rate without sacrificing decoding performance, PC-SIM selects more reliable unfrozen positions for IM and adds inactive information bits (IIBs) to offset rate losses. Additionally, Walsh-Hadamard Transform (WHT) is incorporated to lower the high peak-to-average power ratio (PAPR) in multi-carrier systems and reduce interference. At the receiver, PC-SIM performs polar decoding followed by repetition decoding to obtain index bits and information bits. Simulation results indicate that in Orthogonal Frequency Division Multiplexing (OFDM) systems, compared to conventional polar codes and existing IM-aided polar coding schemes, the proposed PC-SIM scheme significantly improves error performance, coding rate, and PAPR reduction. The proposed PC-SIM achieves around 0.3 dB over the conventional CRC-aided polar codes and IM-aided polar codes with higher coding rate at the bit error ratio (BER) of <inline-formula><tex-math>$4times 10^{-4}$</tex-math></inline-formula>.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2143-2154"},"PeriodicalIF":4.8,"publicationDate":"2025-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11105420","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144842977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the rapid advancement of Vehicle-to-Everything (V2X) technology, there is a growing demand for collaborative perception among vehicles and multimodal devices (e.g., roadside units, pedestrian terminals). However, traditional centralized learning and federated learning (FL) face challenges in model convergence and performance degradation due to non-IID data distribution, privacy protection requirements, and communication bandwidth constraints among massive heterogeneous devices in V2X scenarios. To address these issues, this paper proposes a heterogeneous federated learning framework based on feature prototype alignment and generative knowledge transfer, enabling efficient and secure cross-device collaborative learning. The framework employs dynamic edge-enhanced contrastive learning on the server side to generate trainable global feature prototypes. These prototypes are subsequently decoded into composite images through a pre-trained generative adversarial network, achieving lightweight privacy-preserving knowledge transfer. Experimental results on CIFAR-10, CIFAR-100, and BelgiumTSC datasets demonstrate that our method achieves significant accuracy improvements compared with baseline approaches such as FedDistill and FedTGP. This study establishes a novel theoretical framework and technical pathway for collaborative learning in V2X environments that effectively balances privacy protection with model performance.
{"title":"Heterogeneous Federated Learning for Vehicle-to-Everything: Feature Prototype Aggregation and Generative Feedback Mechanism","authors":"Xianhui Liu;Jianle Liu;Yingyao Zhang;Ning Jia;Chenlin Zhu","doi":"10.1109/OJVT.2025.3594030","DOIUrl":"https://doi.org/10.1109/OJVT.2025.3594030","url":null,"abstract":"With the rapid advancement of Vehicle-to-Everything (V2X) technology, there is a growing demand for collaborative perception among vehicles and multimodal devices (e.g., roadside units, pedestrian terminals). However, traditional centralized learning and federated learning (FL) face challenges in model convergence and performance degradation due to non-IID data distribution, privacy protection requirements, and communication bandwidth constraints among massive heterogeneous devices in V2X scenarios. To address these issues, this paper proposes a heterogeneous federated learning framework based on feature prototype alignment and generative knowledge transfer, enabling efficient and secure cross-device collaborative learning. The framework employs dynamic edge-enhanced contrastive learning on the server side to generate trainable global feature prototypes. These prototypes are subsequently decoded into composite images through a pre-trained generative adversarial network, achieving lightweight privacy-preserving knowledge transfer. Experimental results on CIFAR-10, CIFAR-100, and BelgiumTSC datasets demonstrate that our method achieves significant accuracy improvements compared with baseline approaches such as FedDistill and FedTGP. This study establishes a novel theoretical framework and technical pathway for collaborative learning in V2X environments that effectively balances privacy protection with model performance.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":"6 ","pages":"2332-2342"},"PeriodicalIF":4.8,"publicationDate":"2025-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11103507","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145011313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：