Underwater Multi-Vehicle Navigation by Relative Positioning with LSTM Architecture

This paper introduces a novel AI-based approach to enhance formation control in underwater environments. Deploying multiple robotic vehicles can significantly improve operations such as mapping, surveying, and inspecting areas of interest. These tasks often require maintaining a stable and coordinated formation among the vehicles to ensure efficiency and safety. However, navigating multiple vehicles underwater presents significant challenges, primarily due to the absence of GPS (Global Positioning System) and GNSS (Global Navigation Satellite System) capabilities, which are unavailable underwater. Instead, acoustic solutions and Inertial Navigation Systems (INS) are commonly used, but both come with inherent limitations. Acoustic positioning systems can offer precise localization but rely on costly and logistically demanding infrastructure, such as deploying and synchronizing transponders on the seafloor. On the other hand, INS provides autonomy but suffers from cumulative errors, such as random walk, which degrade accuracy over time. This degradation can lead to task failure or collisions in multi-vehicle formations, especially when operating in dynamic or unstructured environments. To address these challenges, this work proposes a novel approach using the Long Short-Term Memory (LSTM) architecture to mitigate the positional drift of INS while avoiding the logistical complexity of acoustic solutions. The proposed method enhances state estimation and ensures robust formation control in leader-follower setups. This AI-driven solution is scalable and adaptable, offering a significant step forward in multi-vehicle underwater navigation systems.