Parallelized Active Information Gathering Using Multisensor Network for Environment Monitoring
Parallelized Active Information Gathering Using Multisensor Network for Environment Monitoring
复制标题
DOI：
10.1109/tcst.2021.3069175
复制
发表时间：
2021-04
影响因子：
4.8
通讯作者：
Bin Du;Kun Qian;C. Claudel;Dengfeng Sun
中科院分区：
文献类型：
--
作者：
Bin Du;Kun Qian;C. Claudel;Dengfeng Sun
本文以环境监测为应用背景，研究了一种新的算法框架，以并行方式解决多传感器主动信息采集问题。与依赖于完全连接的传感器网络和顺序处理的现有方法不同，我们的方法建立在通用网络拓扑结构上，使各个传感器能够通过与近邻通信来同时做出决策。利用多传感器网络之间的合作，我们表明，所提出的并行化算法的计算复杂度可以大大降低相比，顺序更新计划，同时，所获得的解决方案的次优性得到保证。算法的复杂性和解决方案的质量之间的互连是明确建立相对于网络拓扑结构。在此基础上，进一步给出了在给定算法执行时间预算的情况下设计最优传感器网络的方法。最后，数值模拟的甲烷排放监测方案，以验证我们的方法的有效性。
Motivated by the application of environment monitoring, this article studies a novel algorithmic framework for solving, in a parallelized manner, the problem of multisensor active information gathering. Unlike the existing methods relying on the fully connected sensor network and sequential processing, our approach builds on the generic network topology and enables individual sensors to simultaneously make their decisions by communicating with the immediate neighbors. Leveraging the cooperation among the multisensor network, we show that the computational complexity of the proposed parallelized algorithm can be greatly reduced compared to the sequential updating schemes, and meanwhile, the suboptimality of obtained solutions is guaranteed. The interconnection between the algorithm complexity and solution quality is explicitly established with respect to the network topology. Based on such interconnection, we further provide the approach to design the optimal sensor network by the given time budget of the algorithm execution. Finally, numerical simulations on a methane emission monitoring scenario are presented to validate the effectiveness of our approach.