AQUA-SLAM: Tightly-Coupled Underwater Acoustic-Visual-Inertial SLAM with Sensor Calibration

TL;DR

AQUA-SLAM delivers a tightly-coupled graph-based underwater SLAM that fuses Doppler Velocity Log (DVL) data with stereo visual and inertial measurements, supported by a fast online calibration framework for extrinsics and DVL transducer alignment. The system derives rigorous DVL measurement and pre-integration models, integrates them into an ORB-SLAM3-based pipeline, and introduces rapid linear approximation techniques to enable real-time online calibration. Extensive tank experiments with ground truth and offshore validation in the North Sea demonstrate superior localization accuracy and robustness over state-of-the-art underwater and visual-inertial SLAM systems. The work also provides open-source code to accelerate adoption and further research in underwater localization and mapping.

Abstract

Underwater environments pose significant challenges for visual Simultaneous Localization and Mapping (SLAM) systems due to limited visibility, inadequate illumination, and sporadic loss of structural features in images. Addressing these challenges, this paper introduces a novel, tightly-coupled Acoustic-Visual-Inertial SLAM approach, termed AQUA-SLAM, to fuse a Doppler Velocity Log (DVL), a stereo camera, and an Inertial Measurement Unit (IMU) within a graph optimization framework. Moreover, we propose an efficient sensor calibration technique, encompassing multi-sensor extrinsic calibration (among the DVL, camera and IMU) and DVL transducer misalignment calibration, with a fast linear approximation procedure for real-time online execution. The proposed methods are extensively evaluated in a tank environment with ground truth, and validated for offshore applications in the North Sea. The results demonstrate that our method surpasses current state-of-the-art underwater and visual-inertial SLAM systems in terms of localization accuracy and robustness. The proposed system will be made open-source for the community.

Figures (21)

• Figure 1: Estimated trajectory () and dense 3D reconstruction of an offshore structure using the proposed AQUA-SLAM algorithm. Images (a-d) show the challenging underwater conditions for a SLAM system using a camera.
• Figure 2: Coordinate frames. The world frame $\mathtt{W}$'s z-axis is aligned with the gravity vector. The DVL frame $\mathtt{D}$, the IMU frame $\mathtt{I}$ and the camera frame $\mathtt{C}$ are rigidly fixed on the robot. The DVL sensor measures linear velocity with respect to the seabed. 3D visual landmarks $\mathcal{L}$ are estimated from scenes.
• Figure 3: Sensor measurements from camera, IMU and DVL. Dash yellow triangles represent constant velocity taken from the last DVL measurement.
• Figure 4: DVL transducer measurements with 2D and 3D views. The DVL has 4 transducers facing different directions. Transducer 4 is shown as an example.
• Figure 5: Overview of system implementation.