Pub Date : 2024-07-30DOI: 10.1109/TNSE.2024.3435839
Xiaofeng Liu;Chenqi Guo;Mingjun Zhao;Yinglong Ma
Numerous explainability techniques have been developed to reveal the prediction principles of Graph Neural Networks (GNNs) across diverse domains. However, many existing approaches, particularly those concentrating on model-level explanations, tend to grapple with the tunnel vision problem, leading to less-than-optimal outcomes and constraining users' comprehensive understanding of GNNs. Furthermore, these methods typically require hyperparameters to mold the explanations, introducing unintended human biases. In response, we present GAXG, a global and self-adaptive optimal graph topology generation framework for explaining GNNs' prediction principles at model-level. GAXG addresses the challenges of tunnel vision and hyperparameter reliance by integrating a strategically tailored Monte Carlo Tree Search (MCTS) algorithm. Notably, our tailored MCTS algorithm is modified to incorporate an Edge Mask Learning and Simulated Annealing-based subgraph screening strategy during the expansion phase, resolving the inherent time-consuming challenges of the tailored MCTS and enhancing the quality of the generated explanatory graph topologies. Experimental results underscore GAXG's effectiveness in discovering global explanations for GNNs, outperforming leading explainers on most evaluation metrics.
{"title":"GAXG: A Global and Self-Adaptive Optimal Graph Topology Generation Framework for Explaining Graph Neural Networks","authors":"Xiaofeng Liu;Chenqi Guo;Mingjun Zhao;Yinglong Ma","doi":"10.1109/TNSE.2024.3435839","DOIUrl":"10.1109/TNSE.2024.3435839","url":null,"abstract":"Numerous explainability techniques have been developed to reveal the prediction principles of Graph Neural Networks (GNNs) across diverse domains. However, many existing approaches, particularly those concentrating on model-level explanations, tend to grapple with the tunnel vision problem, leading to less-than-optimal outcomes and constraining users' comprehensive understanding of GNNs. Furthermore, these methods typically require hyperparameters to mold the explanations, introducing unintended human biases. In response, we present GAXG, a global and self-adaptive optimal graph topology generation framework for explaining GNNs' prediction principles at model-level. GAXG addresses the challenges of tunnel vision and hyperparameter reliance by integrating a strategically tailored Monte Carlo Tree Search (MCTS) algorithm. Notably, our tailored MCTS algorithm is modified to incorporate an Edge Mask Learning and Simulated Annealing-based subgraph screening strategy during the expansion phase, resolving the inherent time-consuming challenges of the tailored MCTS and enhancing the quality of the generated explanatory graph topologies. Experimental results underscore GAXG's effectiveness in discovering global explanations for GNNs, outperforming leading explainers on most evaluation metrics.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 6","pages":"6007-6023"},"PeriodicalIF":6.7,"publicationDate":"2024-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141866352","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 : 2024-07-29DOI: 10.1109/TNSE.2024.3433392
Meirong Wang;Jianqiang Hu;Ahmed Alsaedi;Jinde Cao
This paper studies the distributed leader-following consensus problem of unknown nonlinear multi-agent systems (MASs) under false data injection attacks (FDIAs), where the followers connected to the leader may receive the injected false data from the leader's communication channels. Due to the existence of FDIAs, the real and broken leader state value is not available to the followers and cannot be used by followers' controllers, thus an attack compensator based on the errors between the predictive value and the actual measured value is added to the controller to mitigate the adverse effects of attacks. Fuzzy logic systems (FLSs) and Neural Network (NN) techniques are applied to approximate the unknown nonlinear dynamic by estimating the weight matrix. The proposed controller combines attack compensation with unknown nonlinear function compensation, and finally obtains sufficient conditions for the MASs to be ultimately uniformly bounded (UUB). Two algorithms are presented for undirected and directed communication topologies respectively and the simulation results verify the feasibility of the proposed consensus algorithms.
本文研究了未知非线性多Agent系统（MAS）在虚假数据注入攻击（FDIAs）下的分布式领导者-跟随者共识问题，在这种情况下，与领导者相连的跟随者可能会从领导者的通信信道接收到注入的虚假数据。由于 FDIAs 的存在，跟随者无法获得真实的、被破坏的领导者状态值，跟随者的控制器也无法使用，因此需要在控制器中加入一个基于预测值与实际测量值之间误差的攻击补偿器，以减轻攻击的不利影响。模糊逻辑系统（FLS）和神经网络（NN）技术通过估计权重矩阵来近似未知的非线性动态。所提出的控制器将攻击补偿与未知非线性函数补偿相结合，最终获得了 MAS 最终均匀有界（UUB）的充分条件。针对无向和有向通信拓扑分别提出了两种算法，仿真结果验证了所提共识算法的可行性。
{"title":"Leader-Following Consensus Control of Unknown Nonlinear MASs Under False Data Injection Attacks","authors":"Meirong Wang;Jianqiang Hu;Ahmed Alsaedi;Jinde Cao","doi":"10.1109/TNSE.2024.3433392","DOIUrl":"10.1109/TNSE.2024.3433392","url":null,"abstract":"This paper studies the distributed leader-following consensus problem of unknown nonlinear multi-agent systems (MASs) under false data injection attacks (FDIAs), where the followers connected to the leader may receive the injected false data from the leader's communication channels. Due to the existence of FDIAs, the real and broken leader state value is not available to the followers and cannot be used by followers' controllers, thus an attack compensator based on the errors between the predictive value and the actual measured value is added to the controller to mitigate the adverse effects of attacks. Fuzzy logic systems (FLSs) and Neural Network (NN) techniques are applied to approximate the unknown nonlinear dynamic by estimating the weight matrix. The proposed controller combines attack compensation with unknown nonlinear function compensation, and finally obtains sufficient conditions for the MASs to be ultimately uniformly bounded (UUB). Two algorithms are presented for undirected and directed communication topologies respectively and the simulation results verify the feasibility of the proposed consensus algorithms.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 5","pages":"4513-4524"},"PeriodicalIF":6.7,"publicationDate":"2024-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141866357","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 : 2024-07-29DOI: 10.1109/tnse.2024.3433604
Zhenwei Liu, Meirong Zhang, Ali Saberi, Anton A. Stoorvogel
{"title":"Scale-Free Collaborative Protocol Design for Exact Output Synchronization of Multi-Agent Systems in the Presence of Disturbances and Measurement Noise With Known Frequencies","authors":"Zhenwei Liu, Meirong Zhang, Ali Saberi, Anton A. Stoorvogel","doi":"10.1109/tnse.2024.3433604","DOIUrl":"https://doi.org/10.1109/tnse.2024.3433604","url":null,"abstract":"","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"19 1","pages":""},"PeriodicalIF":6.6,"publicationDate":"2024-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141866360","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 : 2024-07-29DOI: 10.1109/tnse.2024.3434957
Yangyang Qian, Zongli Lin, Yacov A. Shamash
{"title":"Fully Distributed Adaptive Resilient Control of Networked Heterogeneous Battery Systems with Unknown Parameters","authors":"Yangyang Qian, Zongli Lin, Yacov A. Shamash","doi":"10.1109/tnse.2024.3434957","DOIUrl":"https://doi.org/10.1109/tnse.2024.3434957","url":null,"abstract":"","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"42 1","pages":""},"PeriodicalIF":6.6,"publicationDate":"2024-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141866355","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}
In recent years, mobile data has grown explosively due to the rapid development of Internet of Vehicles (IoV). However, resources of IoV are limited, in order to alleviate the problem of resource shortage, it is necessary to combine the resource rich aerial cloud and the ground edge nodes. In order to improve efficiency of proactive cache, we propose a proactive cache decision algorithm based on prior knowledge and aerial cloud assistance. Firstly, we divide requests into two types: content download requests and task calculation requests. Then the dynamic request graph based on relationship between users and requests is constructed, temporal graph network and long short term memory are used to predict prior information and caching benefit function is proposed based on popularity and supplemented by prior information to indicate cache location of request content. Finally, the problem of maximizing cache benefit is proposed and the theoretical solution is obtained using Lagrange multiplier method as well as simulation solution is obtained based on Deep Deterministic Policy Gradient. The simulation results demonstrate that the proposed caching scheme can greatly improve caching efficiency, reduce latency and energy consumption.Compared to D3QN, Dueling DQN, and Double DQN, system revenue of proposed algorithm has increased by 66.65%, 177.71% and 36.08%.
{"title":"A Novel Proactive Cache Decision Algorithm Based on Prior Knowledge and Aerial Cloud Assistance in Internet of Vehicles","authors":"Geng Chen;Jingli Sun;Yuxiang Zhou;Qingtian Zeng;Fei Shen","doi":"10.1109/TNSE.2024.3433544","DOIUrl":"10.1109/TNSE.2024.3433544","url":null,"abstract":"In recent years, mobile data has grown explosively due to the rapid development of Internet of Vehicles (IoV). However, resources of IoV are limited, in order to alleviate the problem of resource shortage, it is necessary to combine the resource rich aerial cloud and the ground edge nodes. In order to improve efficiency of proactive cache, we propose a proactive cache decision algorithm based on prior knowledge and aerial cloud assistance. Firstly, we divide requests into two types: content download requests and task calculation requests. Then the dynamic request graph based on relationship between users and requests is constructed, temporal graph network and long short term memory are used to predict prior information and caching benefit function is proposed based on popularity and supplemented by prior information to indicate cache location of request content. Finally, the problem of maximizing cache benefit is proposed and the theoretical solution is obtained using Lagrange multiplier method as well as simulation solution is obtained based on Deep Deterministic Policy Gradient. The simulation results demonstrate that the proposed caching scheme can greatly improve caching efficiency, reduce latency and energy consumption.Compared to D3QN, Dueling DQN, and Double DQN, system revenue of proposed algorithm has increased by 66.65%, 177.71% and 36.08%.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 6","pages":"5280-5297"},"PeriodicalIF":6.7,"publicationDate":"2024-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141772323","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 : 2024-07-24DOI: 10.1109/TNSE.2024.3432917
Jiashi Gao;Ziwei Wang;Xuetao Wei
Artificial intelligence (AI) model services offer remarkable efficiency and automation, engaging customers across various tasks. However, not all AI consumers possess sufficient data to drive AI model training or the specialized knowledge to construct high-performance AI model structures; this has led to a trend in AI model service transactions, a novel facet of the digital economy. Unlike conventional digital products, AI models undergo performance degradation over time. This phenomenon occurs as the training data becomes outdated, leading to a “distribution shift” away from the target distribution of the most recent downstream tasks. This degradation decreases consumer demand, making the AI model less competitive and lowering provider revenue. In this work, we analyze the impact of performance degradation on consumers' demand for AI model services and propose an adaptive pricing framework for service providers to maximize revenue in real-time AI model service exchange. Specifically, We propose an optimal transport (OT) distance-based approach to estimate model performance degradation effectively. Building on this methodology, we implement several practical solutions for predicting changes in future demand rates resulting from current pricing configurations. We then propose a demand-driven AI model update mechanism for service providers to maintain high product demand rates while reducing retraining AI models' costs. We finally propose a reinforcement learning-based pricing mechanism that facilitates adaptive and rapid pricing responses to achieve revenue maximization. Extensive experiments in both 2-competitor and multi-competitor markets validate our framework, showing a significant revenue advantage over baseline pricing strategies in AI model service transactions.
{"title":"An Adaptive Pricing Framework for Real-Time AI Model Service Exchange","authors":"Jiashi Gao;Ziwei Wang;Xuetao Wei","doi":"10.1109/TNSE.2024.3432917","DOIUrl":"10.1109/TNSE.2024.3432917","url":null,"abstract":"Artificial intelligence (AI) model services offer remarkable efficiency and automation, engaging customers across various tasks. However, not all AI consumers possess sufficient data to drive AI model training or the specialized knowledge to construct high-performance AI model structures; this has led to a trend in AI model service transactions, a novel facet of the digital economy. Unlike conventional digital products, AI models undergo performance degradation over time. This phenomenon occurs as the training data becomes outdated, leading to a “distribution shift” away from the target distribution of the most recent downstream tasks. This degradation decreases consumer demand, making the AI model less competitive and lowering provider revenue. In this work, we analyze the impact of performance degradation on consumers' demand for AI model services and propose an adaptive pricing framework for service providers to maximize revenue in real-time AI model service exchange. Specifically, We propose an optimal transport (OT) distance-based approach to estimate model performance degradation effectively. Building on this methodology, we implement several practical solutions for predicting changes in future demand rates resulting from current pricing configurations. We then propose a demand-driven AI model update mechanism for service providers to maintain high product demand rates while reducing retraining AI models' costs. We finally propose a reinforcement learning-based pricing mechanism that facilitates adaptive and rapid pricing responses to achieve revenue maximization. Extensive experiments in both 2-competitor and multi-competitor markets validate our framework, showing a significant revenue advantage over baseline pricing strategies in AI model service transactions.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 5","pages":"5114-5129"},"PeriodicalIF":6.7,"publicationDate":"2024-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141772325","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}
In many web applications, such as Content Delivery Networks (CDNs), TLS credentials are shared, e.g., between the website's TLS origin server and the CDN's edge servers, which can be distributed around the globe. To enhance the security and trust for TLS 1.3 in such scenarios, we propose LURK-T, a provably secure framework which allows for limited use of remote keys with added trust in TLS 1.3. We efficiently decouple the server side of TLS 1.3 into a LURK-T Crypto Service ( $mathit {CS}$ ) and a LURK-T Engine ( $mathit {E}$ ). $mathit {CS}$ executes all cryptographic operations in a Trusted Execution Environment (TEE), upon $mathit {E}$ ’s requests. $mathit {CS}$ and $mathit {E}$ together provide the whole TLS-server functionality. A major benefit of our construction is that it is application agnostic; the LURK-T Crypto Service could be collocated with the LURK-T Engine, or it could run on different machines. Thus, our design allows for in situ attestation and protection of the cryptographic side of the TLS server, as well as for all setups of CDNs over TLS. To support such a generic decoupling, we provide a full Application Programming Interface (API) for LURK-T. To this end, we implement our LURK-T Crypto Service using Intel SGX and integrate it with OpenSSL. We also test LURK-T's efficiency and show that, from a TLS-client's perspective, HTTPS servers using LURK-T instead a traditional TLS-server have no noticeable overhead when serving files greater than 1 MB. In addition, we provide cryptographic proofs and formal security verification using ProVerif.
{"title":"LURK-T: Limited Use of Remote Keys With Added Trust in TLS 1.3","authors":"Behnam Shobiri;Sajjad Pourali;Daniel Migault;Ioana Boureanu;Stere Preda;Mohammad Mannan;Amr Youssef","doi":"10.1109/TNSE.2024.3432836","DOIUrl":"10.1109/TNSE.2024.3432836","url":null,"abstract":"In many web applications, such as Content Delivery Networks (CDNs), TLS credentials are shared, e.g., between the website's TLS origin server and the CDN's edge servers, which can be distributed around the globe. To enhance the security and trust for TLS 1.3 in such scenarios, we propose LURK-T, a provably secure framework which allows for limited use of remote keys with added trust in TLS 1.3. We efficiently decouple the server side of TLS 1.3 into a LURK-T Crypto Service (\u0000<inline-formula><tex-math>$mathit {CS}$</tex-math></inline-formula>\u0000) and a LURK-T Engine (\u0000<inline-formula><tex-math>$mathit {E}$</tex-math></inline-formula>\u0000). \u0000<inline-formula><tex-math>$mathit {CS}$</tex-math></inline-formula>\u0000 executes all cryptographic operations in a Trusted Execution Environment (TEE), upon \u0000<inline-formula><tex-math>$mathit {E}$</tex-math></inline-formula>\u0000’s requests. \u0000<inline-formula><tex-math>$mathit {CS}$</tex-math></inline-formula>\u0000 and \u0000<inline-formula><tex-math>$mathit {E}$</tex-math></inline-formula>\u0000 together provide the whole TLS-server functionality. A major benefit of our construction is that it is application agnostic; the LURK-T Crypto Service could be collocated with the LURK-T Engine, or it could run on different machines. Thus, our design allows for in situ attestation and protection of the cryptographic side of the TLS server, as well as for all setups of CDNs over TLS. To support such a generic decoupling, we provide a full Application Programming Interface (API) for LURK-T. To this end, we implement our LURK-T Crypto Service using Intel SGX and integrate it with OpenSSL. We also test LURK-T's efficiency and show that, from a TLS-client's perspective, HTTPS servers using LURK-T instead a traditional TLS-server have no noticeable overhead when serving files greater than 1 MB. In addition, we provide cryptographic proofs and formal security verification using ProVerif.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 6","pages":"6313-6327"},"PeriodicalIF":6.7,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141772324","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 : 2024-07-23DOI: 10.1109/TNSE.2024.3432746
Yi Zhou;Zhanqi Jin;Huaguang Shi;Lei Shi;Ning Lu;Mianxiong Dong
Cellular networks are difficult to meet emergency rescue due to the destruction of base stations and infrastructure caused by natural disasters. Unmanned Ground Vehicles (UGVs) and other mobile communication devices encounter significant challenges when operating in disaster areas due to limited coverage and resources. To tackle this problem, this paper integrates Unmanned Aerial Vehicles (UAVs) into the emergency communication network and constructs an air-ground integration network architecture with UAV-UGV collaboration. Specifically, multi-UGV collaborate to collect disaster information, and multi-aerial intelligent reflecting surfaces with high maneuverability can effectively assist UGVs in transmitting the collected data to the remote control center. However, there is also a serious challenge to optimize the collaboration strategy between UGVs and UAVs. To address the concern, the collaboration between UAVs and UGVs is modeled as bipartite graph, where UAVs and UGVs belong to different sets of nodes, respectively. The problem is transformed into a matching game based on the bipartite graph. A stable Bidirectional Matching Game (BMG) algorithm is proposed, where matching players maximize the utility by adjusting the selection strategy. Extensive experimental results show that the proposed BMG algorithm outperforms other benchmark algorithms in terms of utility for both UAVs and UGVs.
{"title":"Enhanced Emergency Communication Services for Post–Disaster Rescue: Multi-IRS Assisted Air-Ground Integrated Data Collection","authors":"Yi Zhou;Zhanqi Jin;Huaguang Shi;Lei Shi;Ning Lu;Mianxiong Dong","doi":"10.1109/TNSE.2024.3432746","DOIUrl":"10.1109/TNSE.2024.3432746","url":null,"abstract":"Cellular networks are difficult to meet emergency rescue due to the destruction of base stations and infrastructure caused by natural disasters. Unmanned Ground Vehicles (UGVs) and other mobile communication devices encounter significant challenges when operating in disaster areas due to limited coverage and resources. To tackle this problem, this paper integrates Unmanned Aerial Vehicles (UAVs) into the emergency communication network and constructs an air-ground integration network architecture with UAV-UGV collaboration. Specifically, multi-UGV collaborate to collect disaster information, and multi-aerial intelligent reflecting surfaces with high maneuverability can effectively assist UGVs in transmitting the collected data to the remote control center. However, there is also a serious challenge to optimize the collaboration strategy between UGVs and UAVs. To address the concern, the collaboration between UAVs and UGVs is modeled as bipartite graph, where UAVs and UGVs belong to different sets of nodes, respectively. The problem is transformed into a matching game based on the bipartite graph. A stable Bidirectional Matching Game (BMG) algorithm is proposed, where matching players maximize the utility by adjusting the selection strategy. Extensive experimental results show that the proposed BMG algorithm outperforms other benchmark algorithms in terms of utility for both UAVs and UGVs.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 5","pages":"4651-4664"},"PeriodicalIF":6.7,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141772381","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}
The proliferation of computation-intensive and delay-sensitive services in intelligent transportation systems, such as autonomous driving and vehicle-mounted infotainment services, presents a significant challenge for vehicular users (VUs) with limited resources. To address this issue, multi-access edge computing (MEC) has been considered a favorable solution to mitigate computation delay. This paper considers computation offloading for an air-ground integrated computing platform in vehicular networks. Specifically, we first propose a multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm to optimize the trajectory of UAVs. Then, an algorithm named federated upgraded dueling double deep Q network (FUD3QN) is proposed to meet quality of service (QoS) requirements. The algorithm allocates cross-domain resources after offloading decision-making, aiming to minimize delay and energy consumption while meeting reliability requirements, maximum tolerable delay, communication requirements, and computing limitations. Addressing the non-deterministic polynomial (NP)-hard problem, we employ a multi-agent federated learning and upgraded dueling double deep Q network algorithm (UD3QN) with centralized training and distributed execution. Simulation results illustrate that the MATD3-FUD3QN algorithm proposed significantly surpasses the baselines, highlighting the advantages of introducing UAVs to enhance transmission quality.
{"title":"Hybrid Multi-Server Computation Offloading in Air–Ground Vehicular Networks Empowered by Federated Deep Reinforcement Learning","authors":"Xiaoqin Song;Quan Chen;Shumo Wang;Tiecheng Song;Lei Xu","doi":"10.1109/TNSE.2024.3432765","DOIUrl":"10.1109/TNSE.2024.3432765","url":null,"abstract":"The proliferation of computation-intensive and delay-sensitive services in intelligent transportation systems, such as autonomous driving and vehicle-mounted infotainment services, presents a significant challenge for vehicular users (VUs) with limited resources. To address this issue, multi-access edge computing (MEC) has been considered a favorable solution to mitigate computation delay. This paper considers computation offloading for an air-ground integrated computing platform in vehicular networks. Specifically, we first propose a multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm to optimize the trajectory of UAVs. Then, an algorithm named federated upgraded dueling double deep Q network (FUD3QN) is proposed to meet quality of service (QoS) requirements. The algorithm allocates cross-domain resources after offloading decision-making, aiming to minimize delay and energy consumption while meeting reliability requirements, maximum tolerable delay, communication requirements, and computing limitations. Addressing the non-deterministic polynomial (NP)-hard problem, we employ a multi-agent federated learning and upgraded dueling double deep Q network algorithm (UD3QN) with centralized training and distributed execution. Simulation results illustrate that the MATD3-FUD3QN algorithm proposed significantly surpasses the baselines, highlighting the advantages of introducing UAVs to enhance transmission quality.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"11 6","pages":"5175-5189"},"PeriodicalIF":6.7,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141772322","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：