Optimizing Energy Efficiency in Underwater Acoustic Networks Through Machine Learning Classifiers

Among many challenges in establishing an Underwater Wireless Sensor Network, is the challenge of resource constraints, battery and bandwidth being limited which renders acoustic networks limited in life and application. One identified application of TinyML is the potential of cutting the demand for network resources on the Internet of Things. Based on this hypothesis, this paper attempts to quantify the potential in using machine learning algorithms at the edge of the underwater network to reduce the burden on the battery powered acoustic node through an example automated of pipeline corrosion detection by transmitting only extracted conclusions from data.