Graduate School of Information Science and Technology, Osaka University, 1-5, Yamadaoka, Suita 565-0871, Osaka, Japan.
Graduate School of Economics, Osaka University, 1-7, Machikaneyama-cho, Toyonaka 560-0043, Osaka, Japan.
Sensors (Basel). 2018 Apr 8;18(4):1133. doi: 10.3390/s18041133.
Virtualization of wireless sensor networks (WSN) is widely considered as a foundational block of edge/fog computing, which is a key technology that can help realize next-generation Internet of things (IoT) networks. In such scenarios, multiple IoT devices and service modules will be virtually deployed and interconnected over the Internet. Moreover, application services are expected to be more sophisticated and complex, thereby increasing the number of modifications required for the construction of network topologies. Therefore, it is imperative to establish a method for constructing a virtualized WSN (VWSN) topology that achieves low latency on information transmission and high resilience against network failures, while keeping the topological construction cost low. In this study, we draw inspiration from inter-modular connectivity in human brain networks, which achieves high performance when dealing with large-scale networks composed of a large number of modules (i.e., regions) and nodes (i.e., neurons). We propose a method for assigning inter-modular links based on a connectivity model observed in the cerebral cortex of the brain, known as the exponential distance rule (EDR) model. We then choose endpoint nodes of these links by controlling inter-modular assortativity, which characterizes the topological connectivity of brain networks. We test our proposed methods using simulation experiments. The results show that the proposed method based on the EDR model can construct a VWSN topology with an optimal combination of communication efficiency, robustness, and construction cost. Regarding the selection of endpoint nodes for the inter-modular links, the results also show that high assortativity enhances the robustness and communication efficiency because of the existence of inter-modular links of two high-degree nodes.
无线传感器网络（WSN）的虚拟化被广泛认为是边缘/雾计算的基础模块，它是一种关键技术，可以帮助实现下一代物联网（IoT）网络。在这种场景中，多个物联网设备和服务模块将通过互联网进行虚拟部署和互联。此外，应用服务预计将更加复杂和复杂，从而增加了构建网络拓扑所需的修改数量。因此，建立一种能够在信息传输中实现低延迟和对网络故障具有高弹性，同时保持拓扑结构构建成本低的虚拟化 WSN（VWSN）拓扑的方法势在必行。在本研究中，我们从人类大脑网络中的模块间连接中汲取灵感，该网络在处理由大量模块（即区域）和节点（即神经元）组成的大规模网络时具有高性能。我们提出了一种基于大脑皮层中观察到的连接模型的模块间链路分配方法，称为指数距离规则（EDR）模型。然后，我们通过控制模块间的同配性来选择这些链路的端点节点，同配性是描述大脑网络拓扑连接的特性。我们使用仿真实验来测试我们提出的方法。结果表明，基于 EDR 模型的方法可以构建具有最佳通信效率、鲁棒性和构建成本组合的 VWSN 拓扑。关于模块间链路的端点节点选择，结果还表明，由于存在两个高度数节点的模块间链路，高同配性增强了鲁棒性和通信效率。
