Navigating Uncertainty: Application of Information Theory for Ocean Exploration Using AUV

Autonomous Underwater Vehicles (AUVs) encounter significant challenges in effectively navigating and mapping underwater environments, primarily due to the uncertainties and limitations associated with onboard sensors. These limitations are compounded by the inherent constraints of AUV endurance when tasked with exploring vast underwater regions. Hence, efficient search strategies are critical for locating objects of interest, such as mines, shipwrecks, or plane debris, in marine environments. To tackle these challenges, this study proposes an innovative search methodology that leverages information theoretic concepts such as Kullback-Leibler (KL) divergence and discrimination gain to intelligently explore areas in presence of poor sensor accuracy. The presented method guides the search by quantifying the expected reduction in uncertainty and improving decision-making based on the prior observations. In essence, the presented framework is designed to enhance search performance in scenarios with low-quality sensors characterized by lowered probabilities of true positives and increased probabilities of false alarms. By leveraging information theory, this study demonstrates that the proposed probabilistic approach outperforms traditional sweep search methods. Monte Carlo simulations across various scenarios are conducted to evaluate detection accuracy, exploration-to-exploitation balance, and the number of moves required for successful mine detection. Results highlight the robustness of the proposed method, particularly in terrains with significant variability, offering a practical solution for real-world applications like mine detection and underwater archaeology.