无人机辅助联邦学习受限于地面用户计算能力与参与积极性。为此,文中提出博弈优化算法,构建多无人机-多用户协同的区块链联邦学习系统,引入区块链激励机制。采用DBSCAN(Density-Based Spatial Clustering of Applications with Noise)...无人机辅助联邦学习受限于地面用户计算能力与参与积极性。为此,文中提出博弈优化算法,构建多无人机-多用户协同的区块链联邦学习系统,引入区块链激励机制。采用DBSCAN(Density-Based Spatial Clustering of Applications with Noise)算法对用户分簇,无人机激励用户上传数据并本地训练,再由选举出的无人机聚合全局模型。构建Stackelberg博弈数据交易模型,无人机(领导者)制定奖励策略,用户(追随者)确定数据上传方案,分别设计双方效用函数(含奖励、能耗等因素)。通过模拟退火算法优化追随者发射功率,将领导者优化问题拆分为三个子问题,分别用粒子群、理论推导和黄金分割法求解最优无人机位置、CPU频率及奖励分配比,最终求得博弈均衡解。仿真验证,该算法可有效提升数据采集效率与联邦学习性能。展开更多
Vehicular fog computing(VFC)has been envisioned as an important application of fog computing in vehicular networks.Parked vehicles with embedded computation resources could be exploited as a supplement for VFC.They co...Vehicular fog computing(VFC)has been envisioned as an important application of fog computing in vehicular networks.Parked vehicles with embedded computation resources could be exploited as a supplement for VFC.They cooperate with fog servers to process offloading requests at the vehicular network edge,leading to a new paradigm called parked vehicle assisted fog computing(PVFC).However,each coin has two sides.There is a follow-up challenging issue in the distributed and trustless computing environment.The centralized computation offloading without tamper-proof audit causes security threats.It could not guard against false-reporting,free-riding behaviors,spoofing attacks and repudiation attacks.Thus,we leverage the blockchain technology to achieve decentralized PVFC.Request posting,workload undertaking,task evaluation and reward assignment are organized and validated automatically through smart contract executions.Network activities in computation offloading become transparent,verifiable and traceable to eliminate security risks.To this end,we introduce network entities and design interactive smart contract operations across them.The optimal smart contract design problem is formulated and solved within the Stackelberg game framework to minimize the total payments for users.Security analysis and extensive numerical results are provided to demonstrate that our scheme has high security and efficiency guarantee.展开更多
在智能电网中,拥有可再生能源发电装备的用户可以与他人进行能源交易,以获取利润。自产能源不足的用户可通过从其他有剩余能源的用户购买所需的能源来满足需求。然而如果参与交易无法为用户带来额外收益,用户就不愿意参与此类交易。为...在智能电网中,拥有可再生能源发电装备的用户可以与他人进行能源交易,以获取利润。自产能源不足的用户可通过从其他有剩余能源的用户购买所需的能源来满足需求。然而如果参与交易无法为用户带来额外收益,用户就不愿意参与此类交易。为了提高能源交易参与者的收益,文中提出了一种新的点对点(peer to peer, P2P)能源交易方法,将能源交易描述为能源产消者和拍卖商之间的非合作博弈。买方根据不同的电价调整购买的能源数量,拍卖者控制博弈,卖方不参与博弈,但最终实现效益最大化,然后证明了存在唯一的博弈均衡,以确定市场能源交易价格和数量。利用区块链技术实现了所提出的能源交易方法,以显示实时P2P能源交易的可行性。仿真结果表明,与现有的两种方法相比,所提出的方法参与者累积效益提高了32%以上,验证了其有效性。展开更多
文摘无人机辅助联邦学习受限于地面用户计算能力与参与积极性。为此,文中提出博弈优化算法,构建多无人机-多用户协同的区块链联邦学习系统,引入区块链激励机制。采用DBSCAN(Density-Based Spatial Clustering of Applications with Noise)算法对用户分簇,无人机激励用户上传数据并本地训练,再由选举出的无人机聚合全局模型。构建Stackelberg博弈数据交易模型,无人机(领导者)制定奖励策略,用户(追随者)确定数据上传方案,分别设计双方效用函数(含奖励、能耗等因素)。通过模拟退火算法优化追随者发射功率,将领导者优化问题拆分为三个子问题,分别用粒子群、理论推导和黄金分割法求解最优无人机位置、CPU频率及奖励分配比,最终求得博弈均衡解。仿真验证,该算法可有效提升数据采集效率与联邦学习性能。
基金supported in part by the National Natural Science Foundation of China(61971148)the Science and Technology Program of Guangdong Province(2015B010129001)+2 种基金the Natural Science Foundation of Guangxi Province(2018GXNSFDA281013)the Foundation for Science and Technology Project of Guilin City(20190214-3)the Key Science and Technology Project of Guangxi(AA18242021)
文摘Vehicular fog computing(VFC)has been envisioned as an important application of fog computing in vehicular networks.Parked vehicles with embedded computation resources could be exploited as a supplement for VFC.They cooperate with fog servers to process offloading requests at the vehicular network edge,leading to a new paradigm called parked vehicle assisted fog computing(PVFC).However,each coin has two sides.There is a follow-up challenging issue in the distributed and trustless computing environment.The centralized computation offloading without tamper-proof audit causes security threats.It could not guard against false-reporting,free-riding behaviors,spoofing attacks and repudiation attacks.Thus,we leverage the blockchain technology to achieve decentralized PVFC.Request posting,workload undertaking,task evaluation and reward assignment are organized and validated automatically through smart contract executions.Network activities in computation offloading become transparent,verifiable and traceable to eliminate security risks.To this end,we introduce network entities and design interactive smart contract operations across them.The optimal smart contract design problem is formulated and solved within the Stackelberg game framework to minimize the total payments for users.Security analysis and extensive numerical results are provided to demonstrate that our scheme has high security and efficiency guarantee.
文摘在智能电网中,拥有可再生能源发电装备的用户可以与他人进行能源交易,以获取利润。自产能源不足的用户可通过从其他有剩余能源的用户购买所需的能源来满足需求。然而如果参与交易无法为用户带来额外收益,用户就不愿意参与此类交易。为了提高能源交易参与者的收益,文中提出了一种新的点对点(peer to peer, P2P)能源交易方法,将能源交易描述为能源产消者和拍卖商之间的非合作博弈。买方根据不同的电价调整购买的能源数量,拍卖者控制博弈,卖方不参与博弈,但最终实现效益最大化,然后证明了存在唯一的博弈均衡,以确定市场能源交易价格和数量。利用区块链技术实现了所提出的能源交易方法,以显示实时P2P能源交易的可行性。仿真结果表明,与现有的两种方法相比,所提出的方法参与者累积效益提高了32%以上,验证了其有效性。
