Big Data Forecasting for Improving Maritime Search Operations

This work presents a big data solution for improving maritime search and rescue. In contrast to the current model-based state of the art, we propose using data direct from surface drifters in the vicinity of the last known contact to predict future motion of the missing person, object, or vessel with a deep neural network. Trained and tested on publicly available data from the DARPA Ocean of Things program, we demonstrate >50% reductions in mean squared error at the 4-hour time horizon vs either linear or ocean model-based trajectory estimation solutions.