Improving QoS for large-scale WSNs

The advancements in information and communication technologies have been triggering an increase in miniaturization and ubiquity, paving the way towards new paradigms in embedded computing systems. Modern embedded systems are enabling a number of smaller, smarter and ubiquitous devices, creating an eagerness for monitoring and controlling everything, everywhere. These facts are pushing forward the design of new Wireless Sensor Network (WSN) infrastructures that will tightly interact with the physical environment, in a ubiquitous and pervasive fashion. However, such cyber-physical systems require a rethinking of the usual computing and networking concepts, and given that these computing entities closely interact with their environment, timeliness is of increasing importance. Nevertheless, many other QoS properties such as scalability, energyefficiency and robustness must also be addressed if these infrastructures are to become a reality. This Thesis addresses the use of standard protocols, particularly IEEE 802.15.4 and ZigBee, combined with commercial technologies as a baseline to enable WSN infrastructures capable of supporting the QoS requirements that future large-scale networked embedded systems will impose. Hence, several architectural solutions (mechanisms, algorithms, protocol add-ons) are hereby proposed to address some of the most prominent QoS challenges, such as timeliness, scalability, robustness and energy-efficiency. Importantly, in order to clearly identify the most prominent QoS challenges and to provide effective QoS solutions with close contact with reality, a hands-on approach is followed throughout this Thesis. Hence, we rely upon two real-world application scenarios (i.e. a Datacentre Monitoring (DM) scenario and a Structural Health Monitoring (SHM) scenario), which were engineered, implemented and deployed in the course of this work, to validate and demonstrate this Thesis’ QoS proposals. This strategy enables a deeper understanding of these infrastructures at a more practical level, and provides the proposals with a real-world application context, showing that these network infrastructures have the potential to be used in real-world cyber-physical applications in the near future, if provided with the necessary QoS management mechanisms. Among the proposals, concerning timeliness, for instance, ZigBee cluster-tree topologies are known for a lack of flexibility in adapting to changes in the traffic or bandwidth requirements at runtime, making these infrastructures not capable of allocating more bandwidth to a set of nodes sensing a particular phenomena, or reducing the latency of a data stream. This Thesis proposes a way of dynamically addressing this problem via a mechanism to re-schedule the clusters’ active periods. Concerning the MAC sub-layer of the IEEE 802.15.4 protocol, in this Thesis we carry out an experimental evaluation of a traffic differentiation mechanism, providing the support of different traffic classes to the legacy protocol. This mechanism is also extended to support intra-cluster communications. In addition to timeliness, this mechanism provides and improvement in terms of energy-efficiency. The IEEE 802.15.4 Guaranteed Time Slot mechanism, missing from most stack implementations, is also