Deployment Strategies for Underwater Sensing and Processing Networks

Underwater wireless sensor networks (UWSN) have been evolving for the last decade. UWSNs are used to study and explore underwater life for the benefit of both humans and underwater creatures. Current applications for UWSNs involve navigation, sea current and coral reef monitoring, habitat monitoring of marine life, fish farming, weather forecasting, and oil leakage detection. Such applications require the collection of large amounts of accurate data across various temporal and spatial scales. Given the severely limited bandwidth and the high propagation delay of the underwater acoustic channel, the transmission of these bulky data using acoustic waves is challenging. As a result, the delivery of data requires a significant amount of time and energy. Moreover, underwater sensor nodes use batteries as their main energy source and are connected by underwater modems that are considered to be energy-hungry devices. Because more power is needed to transmit large volumes of data, power consumption is a major performance metric to be considered. One way to mitigate these challenges is to develop underwater systems that allow in-network data processing to be performed underwater rather than offshore. This can be achieved by deploying underwater processing (computerized) nodes that have the ability to sense, relay, process, and analyze the collected data. Computerized nodes are different from the traditional underwater sensor nodes that are limited to sensing, relaying, and transmitting the gathered data. By applying data processing techniques such as aggregation, fusion, and mining to the collected raw data, either the required result (output) is sent or invaluable bits are removed. In either case, only valuable information is transmitted instead of the entire volume of data. As a result, the bandwidth is used efficiently, reductions are made in the end-to-end delay and the power consumption, and the network lifetime is increased. However, underwater processing nodes are more expensive than typical sensor nodes and have more features in terms of allocated power, memory capacity, and higher feature processing unit. In addition, in order to satisfy some network performance objectives and constraints, the processing nodes deployment is challenging as their locations must be carefully chosen. Therefore, we first developed an optimization framework based on Integer Linear Programming (ILP) for the processing node deployment. Then we proposed several efficient placement algorithms for the processing node deployment for the purpose of solving the ILP in polynomial time. We also investigated the effect of the processing node deployment on the surface-level gateway deployment, and we show their tradeoffs. Finally, to evaluate the performance of the processing node deployment, we applied the new technique on a real application and compared it with the existing one