https://www.mdu.se/

Wireless Sensor Networks (WSNs) are interconnections of spatially distributed sensor nodes with low-power wireless communication. In comparison with traditional wireless networks, WSNs provide novel features in network architecture, which are known as low-cost networking, rapid formation, widespread arrangement, self-organisation, and ease of deployment. For these reasons, many WSNs are utilised and interconnected to form massive Internet-of-Things (IoT) applications in areas such as e-healthcare, industrial automation, and smart cities, to name a few. Nevertheless, WSN networking is affected by dynamic network topology – nodes’ mobility and varying network density, and heterogeneous networks (hetnets) – coexisting radio technologies. In more detail, mobile nodes create and break network connections as they move around; varying network density causes alterations in the routing graph; and coexisting radios is a scenario when multiple WSNs utilising different communication technologies are located within same physical space, where network interoperability is required for non-obtrusive operations between the co-located networks. These networking constraints must be properly managed, otherwise, they will lead to non-deterministic and erratic behavior in overall IoT applications, causing degraded Quality-of-Service (QoS).

In recent literature, several proposals primarily address only one of these aspects; either mobility, density, or coexisting radios. In this thesis, we propose a unified approach to manage dynamic networks and hetnets through the support of network-wide decision-making in the management process. A unified method results in reduction of the overhead cost for system resources and computation complexity with respect to the constrained nature of WSNs, and improvement in decision-making efficiency in network management. The main contributions are modelling of the network management process, proposing an architecture, simulating and implementing the proposed architecture, and evaluating the network performance under network management for WSNs. Performance metrics include latency, throughput, and packet loss rate. Software Defined Networking (SDN), a method for programmable network management, is a suitable solution.

The challenges in design and implementation of network management for WSNs have been experimentally studied in the first paper in the thesis. Further, mobility management has been modelled in the second paper with the modular architecture in the third paper. In the third and fourth paper, we implement the proposed architecture and evaluate network performance. In the fifth paper, we have analysed the network performance in different networking architectures. For standardised simulation and evaluations, We used the Contiki and Mininet simulators, and Linux networking simulators, as well as Linux networking concepts on virtualisation. Empirically, simulation results show that the mean latency has improved more than 100 times with network management compared with typical methods using distributed routing protocols, under varying density in hetnets. Also, we observed a smooth-line performance in mobility management with close to zero packet losses under nodes’ mobility with sporadic communication patterns.