Pub Date : 2026-02-04DOI: 10.1016/j.vehcom.2026.101009
Thanh Binh Doan, Tien Hoa Nguyen
{"title":"Deep learning-based analytical approach for coverage energy prediction in UAV-based energy scavenging networks","authors":"Thanh Binh Doan, Tien Hoa Nguyen","doi":"10.1016/j.vehcom.2026.101009","DOIUrl":"https://doi.org/10.1016/j.vehcom.2026.101009","url":null,"abstract":"","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"2 1","pages":""},"PeriodicalIF":6.7,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146135149","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}
Pub Date : 2026-02-04DOI: 10.1016/j.vehcom.2026.101007
Mingfeng Huang, Peng Wang, Athanasios V. Vasilakos, Hai Zhong
{"title":"Task Offloading Based on Lightweight Identity Authentication and Genetic Optimization for the Internet of Vehicles","authors":"Mingfeng Huang, Peng Wang, Athanasios V. Vasilakos, Hai Zhong","doi":"10.1016/j.vehcom.2026.101007","DOIUrl":"https://doi.org/10.1016/j.vehcom.2026.101007","url":null,"abstract":"","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"33 1","pages":""},"PeriodicalIF":6.7,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146135150","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}
Pub Date : 2026-01-26DOI: 10.1016/j.vehcom.2026.101005
Md Abdullah Al Sami, Ibrahim Tanvir, Palash Roy, Md Abdur Razzaque, Md Rafiul Hassan, Mohammad Mehedi Hassan
Unmanned aerial vehicles (UAVs) have seen breakthroughs in forming Aerial Edge Computing (AEC), which executes computationally intensive tasks generated by Internet of Things (IoT) devices, thanks to their ease of deployment, especially in scenarios where traditional terrestrial base stations are damaged and unable to process tasks due to natural disasters. However, an AEC faces significant challenges due to the limited battery capacity of UAVs and the need for efficient collaboration among them to execute tasks. Existing studies often overlook fine-grained task prioritization and balanced load distribution across UAVs, leading to inefficiencies in energy usage and service delay. In this paper, we have developed an optimization framework for efficiently offloading computationally intensive IoT tasks in a three-stage Digital Twin-enabled multi-UAV-based AEC network environment, which jointly minimizes service latency and energy consumption while ensuring the expected load distribution among the UAVs. The formulated framework is a Mixed-Integer Nonlinear Programming (MINLP) problem, which is inherently NP-hard. To address this, we design GLEMATO, a scalable GTrXL-assisted MADDPG framework that learns high-quality offloading policies through memory-aware task prioritization and cooperative multi-agent decision-making in dynamic AEC scenarios. In GLEMATO, while the GTrXL model ensures adaptive task prioritization by considering factors such as task generation time, energy budget, and application deadlines, while the MADDPG enables decentralized policy learning through sharing cooperative state–actions among UAVs. The experimental results, carried out on the OpenAI Gym simulator platform, demonstrate that the developed GLEMATO framework reduces average energy consumption and service latency by 21.8% and 23.3%, respectively, and increases the average task completion ratio by up to 20.1% for computationally intensive tasks compared to the state-of-the-art approaches.
{"title":"A Gated Transformer MADDPG Algorithm for Latency and Energy Aware Task Offloading in Digital Twinning Aerial Edge Computing","authors":"Md Abdullah Al Sami, Ibrahim Tanvir, Palash Roy, Md Abdur Razzaque, Md Rafiul Hassan, Mohammad Mehedi Hassan","doi":"10.1016/j.vehcom.2026.101005","DOIUrl":"https://doi.org/10.1016/j.vehcom.2026.101005","url":null,"abstract":"Unmanned aerial vehicles (UAVs) have seen breakthroughs in forming Aerial Edge Computing (AEC), which executes computationally intensive tasks generated by Internet of Things (IoT) devices, thanks to their ease of deployment, especially in scenarios where traditional terrestrial base stations are damaged and unable to process tasks due to natural disasters. However, an AEC faces significant challenges due to the limited battery capacity of UAVs and the need for efficient collaboration among them to execute tasks. Existing studies often overlook fine-grained task prioritization and balanced load distribution across UAVs, leading to inefficiencies in energy usage and service delay. In this paper, we have developed an optimization framework for efficiently offloading computationally intensive IoT tasks in a three-stage Digital Twin-enabled multi-UAV-based AEC network environment, which jointly minimizes service latency and energy consumption while ensuring the expected load distribution among the UAVs. The formulated framework is a Mixed-Integer Nonlinear Programming (MINLP) problem, which is inherently NP-hard. To address this, we design GLEMATO, a scalable GTrXL-assisted MADDPG framework that learns high-quality offloading policies through memory-aware task prioritization and cooperative multi-agent decision-making in dynamic AEC scenarios. In GLEMATO, while the GTrXL model ensures adaptive task prioritization by considering factors such as task generation time, energy budget, and application deadlines, while the MADDPG enables decentralized policy learning through sharing cooperative state–actions among UAVs. The experimental results, carried out on the OpenAI Gym simulator platform, demonstrate that the developed GLEMATO framework reduces average energy consumption and service latency by 21.8% and 23.3%, respectively, and increases the average task completion ratio by up to 20.1% for computationally intensive tasks compared to the state-of-the-art approaches.","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"117 1","pages":""},"PeriodicalIF":6.7,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146048182","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}
Unmanned aerial vehicle (UAV)-assisted vehicular networks have garnered researchers’ attention as a promising solution to the limitations of vehicle-to-everything (V2X) communication, especially in the dynamic and dense urban scenario due to the non-line-of-sight (NLoS) setup, interference and unreliable links. Intelligent reflecting surfaces (IRS) further enhance communication quality by intelligently manipulating wireless signals when integrated with a UAV-assisted vehicular network. Although the IRS also supports simultaneous transmission and reflection (STAR), here, the passive reflection mode is considered alone. However, within such networks, due to dynamic topology, multi-dimensional state space, energy constraints and decentralized data, efficient resource management, power control and task offloading are compromised. Due to the limited adaptability, poor scalability and convergence of classical optimization techniques and deep reinforcement learning (DRL), we have presented a novel framework based on quantum federated reinforcement learning (QFRL) in this article. The suggested system makes effective use of the quantum properties of superposition and entanglement to facilitate decision-making. Task offloading, power allocation, and UAV trajectory are all taken into account when modelling the optimization problem as a Markov decision process (MDP). In order to guarantee privacy and decentralized intelligence while drastically cutting down on convergence time and computational overhead, a Quantum Neural Network (QNN) is used in federated learning (FL). The suggested QFRL framework performs better than the conventional Deep Deterministic Policy Gradient (DDPG) and Federated Reinforcement Learning (FRL) approaches, according to simulation results. In particular, the QFRL scheme outperforms FRL by 10.62% and DDPG by 49.32% in terms of energy efficiency. Additionally, QFRL exhibits better scalability and convergence speed as the number of vehicle terminals and IRS elements increases. A quantum-enhanced learning technique is established in this work as a potent remedy for the next generation of energy-efficient UAV communication networks.
{"title":"Quantum federated reinforcement learning-based energy efficiency optimization for IRS-assisted underlaying UAV communication","authors":"Haneef Khan , Neeraj Joshi , Abdoh Jabbari , Hussein Zangoti , Hussien T. Alrakah , Ishan Budhiraja","doi":"10.1016/j.vehcom.2026.101003","DOIUrl":"10.1016/j.vehcom.2026.101003","url":null,"abstract":"<div><div>Unmanned aerial vehicle (UAV)-assisted vehicular networks have garnered researchers’ attention as a promising solution to the limitations of vehicle-to-everything (V2X) communication, especially in the dynamic and dense urban scenario due to the non-line-of-sight (NLoS) setup, interference and unreliable links. Intelligent reflecting surfaces (IRS) further enhance communication quality by intelligently manipulating wireless signals when integrated with a UAV-assisted vehicular network. Although the IRS also supports simultaneous transmission and reflection (STAR), here, the passive reflection mode is considered alone. However, within such networks, due to dynamic topology, multi-dimensional state space, energy constraints and decentralized data, efficient resource management, power control and task offloading are compromised. Due to the limited adaptability, poor scalability and convergence of classical optimization techniques and deep reinforcement learning (DRL), we have presented a novel framework based on quantum federated reinforcement learning (QFRL) in this article. The suggested system makes effective use of the quantum properties of superposition and entanglement to facilitate decision-making. Task offloading, power allocation, and UAV trajectory are all taken into account when modelling the optimization problem as a Markov decision process (MDP). In order to guarantee privacy and decentralized intelligence while drastically cutting down on convergence time and computational overhead, a Quantum Neural Network (QNN) is used in federated learning (FL). The suggested QFRL framework performs better than the conventional Deep Deterministic Policy Gradient (DDPG) and Federated Reinforcement Learning (FRL) approaches, according to simulation results. In particular, the QFRL scheme outperforms FRL by 10.62% and DDPG by 49.32% in terms of energy efficiency. Additionally, QFRL exhibits better scalability and convergence speed as the number of vehicle terminals and IRS elements increases. A quantum-enhanced learning technique is established in this work as a potent remedy for the next generation of energy-efficient UAV communication networks.</div></div>","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"58 ","pages":"Article 101003"},"PeriodicalIF":6.5,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014991","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}
Pub Date : 2026-01-20DOI: 10.1016/j.vehcom.2026.101004
Xiaocheng Wang , Qiaoni Han , Jia Guo , Guowen Cheng
In recent years, the significant increase in the number of vehicles has led to driving safety, road congestion, and environmental pollution problems, while the connected and autonomous vehicles that form a platoon and are equipped with cooperative adaptive cruise control (CACC) can greatly improve traffic safety and road capacity. However, due to the limitation of communication resources and the high mobility of vehicles, the vehicle-to-vehicle (V2V) communications always suffer from time-varying delays and random packet dropouts, which seriously compromise platoon stability. Hence, in this work, considering the non-ideal V2V communications, firstly, the Smith predictor is introduced into the CACC-based platoon system to compensate for the equivalent communication delay caused by time-varying delays and random packet dropouts. Secondly, the analysis of plant stability provides guidance for the selection of control gains, and the analysis of string stability presents the theoretical minimum inter-vehicle distances under different communication delays. Further, the optimal control algorithm is applied to get the optimal values of control gains, so as to improve control accuracy and reduce energy consumption. Lastly, through comparisons with the CACC-based and model predictive control (MPC)-based counterparts, the simulation results validate the effectiveness of the proposed scheme in reducing inter-vehicle distance and enhancing tracking performance. They further reveal that the proposed scheme achieves a reduction in energy consumption by 34.24% and 22.58% relative to the CACC-based and MPC-based systems, respectively. Moreover, experiments conducted on the Xtark vehicle platform confirm the superior comprehensive performance of the proposed scheme.
{"title":"A compensation scheme for non-ideal V2V communications in autonomous vehicle platoons","authors":"Xiaocheng Wang , Qiaoni Han , Jia Guo , Guowen Cheng","doi":"10.1016/j.vehcom.2026.101004","DOIUrl":"10.1016/j.vehcom.2026.101004","url":null,"abstract":"<div><div>In recent years, the significant increase in the number of vehicles has led to driving safety, road congestion, and environmental pollution problems, while the connected and autonomous vehicles that form a platoon and are equipped with cooperative adaptive cruise control (CACC) can greatly improve traffic safety and road capacity. However, due to the limitation of communication resources and the high mobility of vehicles, the vehicle-to-vehicle (V2V) communications always suffer from time-varying delays and random packet dropouts, which seriously compromise platoon stability. Hence, in this work, considering the non-ideal V2V communications, firstly, the Smith predictor is introduced into the CACC-based platoon system to compensate for the equivalent communication delay caused by time-varying delays and random packet dropouts. Secondly, the analysis of plant stability provides guidance for the selection of control gains, and the analysis of string stability presents the theoretical minimum inter-vehicle distances under different communication delays. Further, the optimal control algorithm is applied to get the optimal values of control gains, so as to improve control accuracy and reduce energy consumption. Lastly, through comparisons with the CACC-based and model predictive control (MPC)-based counterparts, the simulation results validate the effectiveness of the proposed scheme in reducing inter-vehicle distance and enhancing tracking performance. They further reveal that the proposed scheme achieves a reduction in energy consumption by 34.24% and 22.58% relative to the CACC-based and MPC-based systems, respectively. Moreover, experiments conducted on the Xtark vehicle platform confirm the superior comprehensive performance of the proposed scheme.</div></div>","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"58 ","pages":"Article 101004"},"PeriodicalIF":6.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014990","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}
Pub Date : 2026-01-13DOI: 10.1016/j.vehcom.2026.101001
Umar Draz , Tariq Ali , Sana Yasin , Mohammad Hijji , Muhammad Ayaz , Saleh Albelwi
Free-Space Optical (FSO) communication is a promising candidate for high-capacity backhaul in beyond 5G (B5G) and 6G networks due to its exceptional bandwidth efficiency, minimal interference, and elevated data rates. However, its vulnerability to adverse weather conditions–such as fog, rain, and turbulence–poses significant limitations. To overcome these challenges, hybrid FSO/RF architectures have been introduced; however, current implementations typically rely on rigid resource allocation schemes, static channel selection, and suboptimal UAV positioning, which limit their adaptability in dynamic environments. In this study, we introduce the Smart Backhaul Framework for UAV Communication (SBF-UC), an intelligent, simulation-validated architecture designed to enhance hybrid FSO/RF backhauling. The framework employs Multi-Agent Reinforcement Learning (MARL) in conjunction with Matching Game Theory (MGT) to enable UAVs to act as autonomous agents that optimize flight altitude, communication link selection, and bandwidth distribution based on visibility-aware environmental metrics. The hybrid switching mechanism ensures continuous connectivity by dynamically selecting between FSO and RF links in response to atmospheric degradation. Extensive simulations under parameterized meteorological scenarios validate the effectiveness of SBF-UC, achieving up to 88% throughput under 30 dB/km attenuation, 25% energy efficiency gains, and latency below 200 ms for a network of 350 users. It contributes a unified multi-agent framework that combines MARL-driven UAV autonomy with matching-based hybrid FSO/RF backhaul optimization, enabling resilient link switching and efficient resource allocation under dynamic atmospheric conditions.The proposed framework offers a robust, scalable, and adaptive solution for resilient aerial backhauling in next-generation mobile communication systems.
{"title":"Adaptive backhaul optimization with hybrid FSO/RF links using multi-agent intelligence in 6G UAV networks","authors":"Umar Draz , Tariq Ali , Sana Yasin , Mohammad Hijji , Muhammad Ayaz , Saleh Albelwi","doi":"10.1016/j.vehcom.2026.101001","DOIUrl":"10.1016/j.vehcom.2026.101001","url":null,"abstract":"<div><div>Free-Space Optical (FSO) communication is a promising candidate for high-capacity backhaul in beyond 5G (B5G) and 6G networks due to its exceptional bandwidth efficiency, minimal interference, and elevated data rates. However, its vulnerability to adverse weather conditions–such as fog, rain, and turbulence–poses significant limitations. To overcome these challenges, hybrid FSO/RF architectures have been introduced; however, current implementations typically rely on rigid resource allocation schemes, static channel selection, and suboptimal UAV positioning, which limit their adaptability in dynamic environments. In this study, we introduce the Smart Backhaul Framework for UAV Communication (SBF-UC), an intelligent, simulation-validated architecture designed to enhance hybrid FSO/RF backhauling. The framework employs Multi-Agent Reinforcement Learning (MARL) in conjunction with Matching Game Theory (MGT) to enable UAVs to act as autonomous agents that optimize flight altitude, communication link selection, and bandwidth distribution based on visibility-aware environmental metrics. The hybrid switching mechanism ensures continuous connectivity by dynamically selecting between FSO and RF links in response to atmospheric degradation. Extensive simulations under parameterized meteorological scenarios validate the effectiveness of SBF-UC, achieving up to 88% throughput under 30 dB/km attenuation, 25% energy efficiency gains, and latency below 200 ms for a network of 350 users. It contributes a unified multi-agent framework that combines MARL-driven UAV autonomy with matching-based hybrid FSO/RF backhaul optimization, enabling resilient link switching and efficient resource allocation under dynamic atmospheric conditions.The proposed framework offers a robust, scalable, and adaptive solution for resilient aerial backhauling in next-generation mobile communication systems.</div></div>","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"58 ","pages":"Article 101001"},"PeriodicalIF":6.5,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145962603","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}
Pub Date : 2026-01-10DOI: 10.1016/j.vehcom.2026.100999
Sang-Quang Nguyen , Duy Tran Trung , Lam-Thanh Tu , Anh Le-Thi , Mui Van Nguyen
This paper proposes a novel secure downlink framework that integrates Partial Non-Orthogonal Multiple Access (PNOMA) with short-packet communications (SPC) under keyhole fading channels, tailored for ultra-reliable low-latency (URLLC) services. Unlike prior studies that addressed NOMA, SPC, or keyhole effects in isolation, our work is the first to jointly consider all three aspects in a unified design. Closed-form expressions for the average secure block error rate (SBLER) and block error rate (BLER) are derived under both partial and full transmission information (PTI/FTI) assumptions at the eavesdropper, together with asymptotic analysis capturing the impact of blocklength, power allocation, and keyhole severity. Numerical simulations confirm that the proposed PNOMA-SPC system consistently outperforms conventional NOMA scheme in terms of latency, reliability, and secrecy, even under strong eavesdropping conditions. These contributions provide new theoretical and practical insights into the secure design of multiple access schemes for next-generation 6G URLLC scenarios.
{"title":"Securing short-packet transmissions via partial NOMA: Performance analysis under keyhole fading","authors":"Sang-Quang Nguyen , Duy Tran Trung , Lam-Thanh Tu , Anh Le-Thi , Mui Van Nguyen","doi":"10.1016/j.vehcom.2026.100999","DOIUrl":"10.1016/j.vehcom.2026.100999","url":null,"abstract":"<div><div>This paper proposes a novel secure downlink framework that integrates Partial Non-Orthogonal Multiple Access (PNOMA) with short-packet communications (SPC) under keyhole fading channels, tailored for ultra-reliable low-latency (URLLC) services. Unlike prior studies that addressed NOMA, SPC, or keyhole effects in isolation, our work is the first to jointly consider all three aspects in a unified design. Closed-form expressions for the average secure block error rate (SBLER) and block error rate (BLER) are derived under both partial and full transmission information (PTI/FTI) assumptions at the eavesdropper, together with asymptotic analysis capturing the impact of blocklength, power allocation, and keyhole severity. Numerical simulations confirm that the proposed PNOMA-SPC system consistently outperforms conventional NOMA scheme in terms of latency, reliability, and secrecy, even under strong eavesdropping conditions. These contributions provide new theoretical and practical insights into the secure design of multiple access schemes for next-generation 6G URLLC scenarios.</div></div>","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"58 ","pages":"Article 100999"},"PeriodicalIF":6.5,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145957191","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}
Pub Date : 2026-01-06DOI: 10.1016/j.vehcom.2026.101000
Yabin Zhu , Xu Zhao , Xin Zhang
Vehicle-to-Everything (V2X) technology is rapidly developing. However, vehicular devices operate with limited computational power and energy. These constraints pose significant challenges for secure and energy-efficient task offloading. To address these challenges, this paper proposes a novel framework that integrates a Graph Neural Network (GNN) with the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm for secure task offloading and resource allocation. The framework employs a GNN (GraphSAGE) to capture the dynamic network topology and global interference, overcoming the limitations of partial observability. This spatial feature representation supports coordinated decision-making by multiple agents within the MADDPG architecture. To handle the high-dimensional and coupled action space, a combinatorial action selection strategy is proposed and QMIX value function decomposition is adopted. This “optimize-then-combine” mechanism enables efficient joint optimization of continuous resources and discrete decisions. Furthermore, a hybrid RSA-AES encryption scheme combined with frequency hopping is implemented to ensure end-to-end data security and anti-jamming capabilities. Extensive comparative experiments demonstrated that the proposed framework significantly outperformed baseline methods, including DQN and standard MADDPG, in terms of task completion rate, average latency, and energy consumption, especially in high-load scenarios. Ablation studies further validated the critical contributions of the GNN, combinatorial action design, and security mechanisms. This work provides an efficient, secure, and scalable solution for resource optimization in complex V2X environments.
{"title":"A secure GNN-MADDPG framework with combinatorial action optimization for task offloading in vehicular networks","authors":"Yabin Zhu , Xu Zhao , Xin Zhang","doi":"10.1016/j.vehcom.2026.101000","DOIUrl":"10.1016/j.vehcom.2026.101000","url":null,"abstract":"<div><div>Vehicle-to-Everything (V2X) technology is rapidly developing. However, vehicular devices operate with limited computational power and energy. These constraints pose significant challenges for secure and energy-efficient task offloading. To address these challenges, this paper proposes a novel framework that integrates a Graph Neural Network (GNN) with the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm for secure task offloading and resource allocation. The framework employs a GNN (GraphSAGE) to capture the dynamic network topology and global interference, overcoming the limitations of partial observability. This spatial feature representation supports coordinated decision-making by multiple agents within the MADDPG architecture. To handle the high-dimensional and coupled action space, a combinatorial action selection strategy is proposed and QMIX value function decomposition is adopted. This “optimize-then-combine” mechanism enables efficient joint optimization of continuous resources and discrete decisions. Furthermore, a hybrid RSA-AES encryption scheme combined with frequency hopping is implemented to ensure end-to-end data security and anti-jamming capabilities. Extensive comparative experiments demonstrated that the proposed framework significantly outperformed baseline methods, including DQN and standard MADDPG, in terms of task completion rate, average latency, and energy consumption, especially in high-load scenarios. Ablation studies further validated the critical contributions of the GNN, combinatorial action design, and security mechanisms. This work provides an efficient, secure, and scalable solution for resource optimization in complex V2X environments.</div></div>","PeriodicalId":54346,"journal":{"name":"Vehicular Communications","volume":"58 ","pages":"Article 101000"},"PeriodicalIF":6.5,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145902773","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}
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：