Dynamic Clustering and Data Aggregation for the Internet-of-Underwater-Things Networks

Advances in semiconductor technology have made it possible to have high processing powers in cheap microcontrollers, which is spawning off a revolution in the range of applications of the Internet-of-Things (IoT) and its underwater counterpart, the Internet-of-Underwater-Things (IoUT). As a result, it has now become possible and cost effective to implement powerful data processing algorithms on very cheap microcontrollers and achieve network intelligence on edge devices. In this paper, we evaluate the impact of implementing an unsupervised machine learning technique based on the k-means algorithm, as well as data aggregation, on the performance of a wireless underwater sensor network. A clustering algorithm based on the k-means algorithm is used to divide the network into clusters and to select cluster heads based on network topology and residual energy. Each cluster head collects and aggregates data from nodes within its cluster's coverage and forwards the data to the sink. The network is deployed in a shallow seabed, and it is assumed that the nodes can reach the sink using their full transmission powers. Hence, the performance evaluation compares the sum-throughput, energy efficiency and coverage probability for direct transmissions to the sink against transmissions using the cluster heads. We also propose a special consideration for disaster early warning data, which packets are assigned priority delivery and handled with minimum delay. The evaluation is performed through computer simulations and the results show over a 100% improvement in throughput for clusterbased transmissions compared to direct transmissions.