Toward Ocean Model-Driven Robotic Exploration

Abstract

This interdisciplinary work demonstrates the viability of coupling ocean models with in situ robotic sampling in a dynamic coastal region as a means to increase model skill and prediction. Model-driven exploration closes the sample-assimilate-​predict-direct loop within a virtuous cycle to refine prediction for a range of applications in a region characterized by harsh conditions, high variability, and diverse physical forcings, including bathymetry and coastal topography. By exploring the feasibility of coupling high-resolution autonomous underwater vehicle (AUV) sampling and data assimilation with a geostatistical model in a continuous loop, we aim to provide a new approach to understanding coastal dynamics and processes, while using modest computational resources. The novelty of this effort is threefold: the demonstration of coupling models with AUV sampling, the importance of targeted sampling, and the impact of loop closure toward model prediction.

Bibliographic Information