为了提高无人机基站(unmanned aerial vehicle base stations,UAV-BS)为地面多用户服务时的数据速率,提出一种基于决斗深度神经网络(dueling deep Q-network,Dueling-DQN)的深度强化学习(deep reinforcement learning,DRL)算法。采用决...为了提高无人机基站(unmanned aerial vehicle base stations,UAV-BS)为地面多用户服务时的数据速率,提出一种基于决斗深度神经网络(dueling deep Q-network,Dueling-DQN)的深度强化学习(deep reinforcement learning,DRL)算法。采用决斗网络(dueling network,DN)结构以克服动态环境的部分可观测问题,联合优化了UAV-BS的位置和下行链路功率分配,在更符合实际的空地概率信道模型中检验了Dueling-DQN算法的性能。结果表明,相较于对比算法,所提出的Dueling-DQN算法可以提供更高的数据速率和服务公平性,且随着地面用户数量的增大,算法的优势更加明显。Dueling-DQN算法可有效解决复杂非凸性问题,为UAV-BS的资源分配问题提供理论参考。展开更多
In this paper, a two-tiered Wireless Sensor Network (WSN) where nodes are divided into clusters and nodes forward data to base stations through cluster heads is considered. To maximize the network lifetime, two energy...In this paper, a two-tiered Wireless Sensor Network (WSN) where nodes are divided into clusters and nodes forward data to base stations through cluster heads is considered. To maximize the network lifetime, two energy efficient approaches are investigated. We first propose an approach that optimally locates the base stations within the network so that the distance between each cluster head and its closest base station is decreased. Then, a routing technique is developed to arrange the communication between cluster heads toward the base stations in order to guaranty that the gathered information effectively and efficiently reach the application. The overall dynamic framework that combines the above two schemes is described and evaluated. The experimental performance evaluation demonstrates the efficacy of topology control as a vital process to maximize the network lifetime of WSNs.展开更多
文摘为了提高无人机基站(unmanned aerial vehicle base stations,UAV-BS)为地面多用户服务时的数据速率,提出一种基于决斗深度神经网络(dueling deep Q-network,Dueling-DQN)的深度强化学习(deep reinforcement learning,DRL)算法。采用决斗网络(dueling network,DN)结构以克服动态环境的部分可观测问题,联合优化了UAV-BS的位置和下行链路功率分配,在更符合实际的空地概率信道模型中检验了Dueling-DQN算法的性能。结果表明,相较于对比算法,所提出的Dueling-DQN算法可以提供更高的数据速率和服务公平性,且随着地面用户数量的增大,算法的优势更加明显。Dueling-DQN算法可有效解决复杂非凸性问题,为UAV-BS的资源分配问题提供理论参考。
文摘In this paper, a two-tiered Wireless Sensor Network (WSN) where nodes are divided into clusters and nodes forward data to base stations through cluster heads is considered. To maximize the network lifetime, two energy efficient approaches are investigated. We first propose an approach that optimally locates the base stations within the network so that the distance between each cluster head and its closest base station is decreased. Then, a routing technique is developed to arrange the communication between cluster heads toward the base stations in order to guaranty that the gathered information effectively and efficiently reach the application. The overall dynamic framework that combines the above two schemes is described and evaluated. The experimental performance evaluation demonstrates the efficacy of topology control as a vital process to maximize the network lifetime of WSNs.
基金国家自然科学基金(the National Natural Science Foundation of China under Grant No.60173026)教育部科研重点项目(the Research Project of MOE of China under Grant No.105071)上海高校网格技术E-研究院资助(200301- 1)
