Use of Inferential Statistics to Design Effective Communication Protocols for Wireless Sensor Networks

This thesis explores the issues and techniques associated with employing the principles of inferential statistics to design effective Medium Access Control (MAC), routing and duty cycle management strategies for multihop Wireless Sensor Networks (WSNs). The main objective of these protocols are to maximise the throughput of the network, to prolong the lifetime of nodes and to reduce the end-to-end delay of packets over a general network scenario without particular considerations for specific topology configurations, traffic patterns or routing policies. WSNs represent one of the leading-edge technologies that have received substantial research efforts due to their prominent roles in many applications. However, to design effective communication protocols for WSNs is particularly challenging due to the scarce resources of these networks and the requirement for large-scale deployment. The MAC, routing and duty cycle management protocols are amongst the important strategies that are required to ensure correct operations of WSNs. This thesis makes use of the inferential statistics field to design these protocols; inferential statistics was selected as it provides a rich design space with powerful approaches and methods. The MAC protocol proposed in this thesis exploits the statistical characteristics of the Gamma distribution to enable each node to adjust its contention parameters dynamically based on its inference for the channel occupancy. This technique reduces the service time of packets and leverages the throughput by improving the channel utilisation. Reducing the service time minimises the energy consumed in contention to access the channel which in turn prolongs the lifetime of nodes. The proposed duty cycle management scheme uses non-parametric Bayesian inference to enable each node to determine the best times and durations for its sleeping durations without posing overheads on the network. Hence the lifetime of node is prolonged by mitigating the amount of energy wasted in overhearing and idle listening. Prolonging the lifetime of nodes increases the throughput of the network and reduces the end-to-end delay as it allows nodes to route their packets over optimal paths for longer periods. The proposed routing protocol uses one of the state-of-the-art inference techniques dubbed spatial reasoning that enables each node to figure out the spatial relationships between nodes without overwhelming the network with control packets. As a result, the end-to-end delay is reduced while the throughput and lifetime are increased. Besides the proposed protocols, this thesis utilises the analytical aspects of statistics to develop rigorous analytical models that can accurately predict the queuing and medium access delay and energy consumption over multihop networks. Moreover, this thesis provides a broader perspective for design of communication protocols for WSNs by casting the operations of these networks in the domains of the artificial chemistry discipline and the harmony search optimisation algorithm.