Drifting 3D Underwater Wireless Sensor Networks for Pollution Monitoring

<p>Increasing demands by a growing population for food and oil have resulted in synchronous rises in the past 40 years of aquaculture and offshore oil drilling. Growth in these industries has highlighted the potential crippling impacts of coastal pollution, with marine farms likely to be susceptible to damage from harmful algal blooms and oil platforms to cause oil spills, as headlined by the 2010 Deepwater Horizon incident in the Gulf of Mexico. Drifting underwater wireless sensor networks (UWSNs) represent a technology that can greatly enhance our understanding of such processes. Consisting of a swarm of untethered sensor nodes, an UWSN can be deployed over a large segment of the feature. The feature is carried through to different positions by underwater currents, the nodes, also drifting with the currents, are able to follow it and track it. To enable UWSNs, work is proceeding at multiple laboratories throughout the world on the design of localization, medium access control and routing protocols that can adapt to the problem that a drifting network topology has over short time intervals: frequent neighbour changes. However, the long-term impact that current drift has on 3D drifting UWSNs is poorly understood. Current mobility models used to generate the motion of drifting nodes in simulation do not reflect how devices in real-life will disperse in water. At present, UWSN protocol schemes are evaluated in simulations where the current mobility of nodes is generated by unrealistic land-based random waypoint and other stochastic mobility models, or coarsely resolved numerical ocean models. Recently, a physically-inspired current mobility model known as MCM has been proposed. This is only defined along the water's surface, in 2D, however, and 3D extensions of this model have been simplistic and arbitrary. The lack of realistic 3D current mobility models motivates this thesis to develop one so that the simulated evolution of UWSNs can more accurately reflect real-life. A consideration of the oceanographic data on which MCM is based is used to derive a 3D extension of the model that reflects observed features of the Gulf Stream current. The speed of the current declines with depth. This model is utilized to advect a drifting UWSN in an oil plume monitoring scenario and study the performance of two pressure routing protocols, Depth Based Routing (DBR) and HydroCast, over time. Previously, these schemes had only been validated in unrealistic stochastic and depth-invariant mobility models and their time-averaged performances were only reported. Our findings show that 3D UWSNs cannot expect to stay connected and functioning if nodes only passively drift with the currents, with results demonstrating that a fully connected network can be so dispersed after three hours that no paths to sinks exist at all. Nodes must be equipped with some form of mobility to prevent their being separated and carried out of the monitoring region. Autonomous underwater vehicles (AUVs) capable of omnidirectional motion are expensive however, and instead UWSNs consisting of pro ling floats are considered in this thesis. Floats can move up and down by adjusting their buoyancy, which also allow them to move in 3D by exploiting water layers that are flowing in different directions to proceed along a desired heading. Recently, promising results in 2D path planning and formation control of floats have been presented. However, these works consider the floats to have infinite vertical velocity whereas in reality this is around 0:3 metres per second. In this thesis, a practical node movement scheme is proposed for extending the coverage lifetime of a 3D UWSN consisting of floats. By accounting for the finite profiling velocity of floats, the scheme is able to position nodes at coverage holes with greater precision than existing 2D strategies. The scheme's performance is analysed by simulations. The results support the use of floats for achieving partial coverage, which can achieve similar levels of coverage as an AUV strategy while requiring less cost to deploy. In determining the lifetime of the network, we find that both energy and dispersion limit the lifetime of the network. Without propulsion, the nodes are carried out of and cease to cover the monitoring region. Using mobility to remain with the region, at the end of a 5 day mission duration floats have had to use up some or almost all of their battery capacity in profiling.</p>