Search

Over the last 10 years, Wireless Sensor Networks (WSNs) have evolved as a hot interdisciplinary research area. What started as a concept of ubiquitous computing with the initial assumptions of low-cost and low power sensing, has now progressed to the development of new wireless sensing applications supported by the advancement of computing devices. Similarly, a vast number of said applications have also emerged with varying Quality of Service (QoS) requirements. It is our argument that the current approach whereby there is a need to customise protocols for individual applications contributes to the slow uptake WSN in the real world. Likewise, we believe that generalised solutions are not always appropriate and while tailor-made solutions are costly. Therefore, we advocate that protocols that understand the notion of a classification of WSN applications can be a potential solution to this problem; in that the tailoring pertains to the type of application or their behaviour classes. Hence, we present the design of a classification-based cross-layer WSN stack that considers Dual-Mode Periodic Data Collection and Event- Driven Monitoring (D-PEDM) applications. In this thesis, we first derive the D-PEDM protocol configuration parameters through thorough analysis of the QoS requirements, physical constraints, and resource constraints of D-PEDM applications. Once the DPEDM protocol configuration parameters are derived, we design an Emergent Broadcast Slot (EBS) scheme, as well as Medium Access Control (YA-MAC) and efficient Priority Based Routing (PBR) protocols. We evaluate each protocol individually as well as an integrated cross-layer stack using the WSN standard test-beds. Further, we show how an Adaptive Rate Control (ARC) protocol can be used with these types of applications. More specifically we introduce the EBS synchronisation scheme that efficiently handles control messages (broadcast messages) in Duty-Cycled Multi-hop (DCM) networks by integrating routing and application broadcast messages together. This approach minimises the overhead caused by control messages that results in improved application level throughput. Additionally, EBS enables efficient decentralised sleep-awake coordination among neighbourhood nodes that improves the power usage in battery-powered sensor devices up to 5-6 times. To meet the latency requirements of the D-PEDM applications, YA-MAC provides support to handle unicast messages in an asynchronous manner that not only reduces the event traffic latency but also reduces contention in the network. To further improve the latency and reliabilities of bursty event traffic; PBR uses a priority queuing mechanism and traffic differentiation schemes. In the traffic differentiation scheme, PBR gives priority to event traffic and diverts non-priority traffic away from the event traffic path to alternative random paths. This favours the QoS of event traffic in many ways: first, it reduces the contention in the event traffic path resulting in less collisions and fewer back-offs at MAC layer, which in turn significantly improves end-to-end event traffic latency (by 10 ms per hop). Additionally, the MAC layers reduced numbers of back-offs inherently reduces re-transmissions by 70% thereby reducing the energy consumption by 2/3. Furthermore, PBR increases the probability of event data reaching the Sink/Base-station as compared to that of when both event and non-event share same paths. The D-PEDM cross-layer stack, which includes EBS, YA-MAC, and PBR, achieves 80% application level throughput while operating in less than 10% duty-cycle in presence of relatively high unicast data traffic.