MVINS: A Magnetism&vision Aided Inertial Navigation System for Autonomous Underwater Vehicles

We present a robust underwater navigation system that integrates magnetic, visual, and inertial measurements from commercial off-the-shelf sensors. Visual Inertial Navigation Systems (VINS) face challenges when used for Autonomous Underwater Vehicle (AUV) localization in perceptually degraded environments. First, traditional VINS methods struggle to accurately detect sufficient loops due to several factors: feature scarcity, environmental similarities, limited visibility, orientation changes, and constrained computational resources. Second, the yaw is unobservable in VINS and it may drift rapidly without distinct features. To address these issues, we propose a novel system that enhances loop closure by fusing magnetic signatures from a low-cost alternating magnetic field coil with multi-scale mapping and hierarchical place recognition. Additionally, we utilize geomagnetic fields to align feature descriptors, improving robustness to orientation variations. Our system also refines yaw estimations by leveraging geomagnetic data, aligning them with global references to mitigate drift. Experimental results validate the improved performance of the proposed system.