AQUA SLAM

1 Introduction

AQUA-SLAM is an underwater SLAM system that integrates a Doppler Velocity Log (DVL), Inertial Measurement Unit (IMU), and stereo cameras to deliver precise and reliable pose estimation and environmental reconstruction in challenging underwater scenarios.

2 Publication

Shida Xu, Kaicheng Zhang, and Sen Wang. "AQUA-SLAM: Tightly-Coupled Underwater Acoustic-Visual-Inertial SLAM with Sensor Calibration." IEEE Transactions on Robotics, 2025. [IEEE] [PDF]

3 Installation

Tested on Ubuntu20.04 and Ubuntu22.04. Local Host:

mkdir aqua_slam_ws/src -p
cd aqua_slam_ws/src
git clone https://github.com/DaDa0o0/AQUA_SLAM_test.git
cd ./AQUA_SLAM_test/docker
docker compose build
docker compose up -d
xhost +
docker exec -it orb_dvl2_ros_noetic bash

In docker container:

catkin_make -DCMAKE_CXX_COMPILER=/usr/bin/clang++-10 DCMAKE_C_COMPILER=/usr/bin/clang-10
mkdir -p ./src/AQUA_SLAM/Vocabulary
cd ./src/AQUA_SLAM/Vocabulary
/root/.local/bin/gdown --fuzzy https://drive.google.com/file/d/1kFIh_By8wMj6AySCJ1aKjWvi7EHP7O7G/view\?usp\=sharing
cd /root/catkin_ws
source devel/setup.bash
roslaunch ORB_DVL2 blue_gx5_StructureEasy.launch

You should be able to see a rviz window after this step.

Note: We provide multiple launch files for each sequence (e.g., blue_gx5_StructureEasy.launch for Structure_Easy.bag), as some sequences use different extrinsic calibrations. Please make sure to use the corresponding launch file when running each ROS bag.

4 Running with Tank Dataset

Dataset Download

Please visit Tank Dataset website to download ros bags.

Here we download Structure_Easy.bag for test.

Place the ros bag under aqua_slam_ws/src/AQUA_SLAM_test/dataset

Run with a ROS bag

Open a new terminal on local host to play the rosbag:

docker exec -it orb_dvl2_ros_noetic bash
rosbag play ./src/AQUA_SLAM/dataset/Structure_Easy.bag

Now you should see the SLAM running in rviz.

5 Run with your own data

You need to modify the camera parameters and extrinsic parameters to your own setting in the YAML file under data folder.

Important parameters

Here we list some parameter affect performance seriously:

• Sensor Extrinsics:

  • T_dvl_c: Transformation matrix from the DVL frame to the Camera frame.
  • T_gyro_c: Rotation matrix from the IMU frame to the Camera frame.

• IMU & DVL Settings:

  • IMU.NoiseGyro, IMU.NoiseAcc: Noise parameters for the IMU.
  • IMU.GyroWalk, IMU.AccWalk: Random walk parameters for the IMU.
  • IMU.Frequency: Frequency of the IMU.
  • alpha,beta: DVL Transducer orientation.

• Initialization:

  • IMUInitTranslation: The distance threshold for IMU initialization. IMU initialization will be triggered when the distance traveled by the camera exceeds this threshold.
  • IMUInitRotation: The rotation threshold for IMU initialization. IMU initialization will be triggered when the rotation angle of the camera exceeds this threshold.

• Loop Closure

  • EnableLoopDetection: Enable/Disable loop closure detection.

6 Results

7 Known Issues

• Currently, the system may randomly crash when running with long sequences due to some multithreading issues. We are working on fixing this problem. If you encounter a crash, please try restarting the system again.

8 Citation

If you find this work useful in your research, please consider citing:

@article{xu2025aqua,
  title={AQUA-SLAM: Tightly-Coupled Underwater Acoustic-Visual-Inertial SLAM with Sensor Calibration},
  author={Xu, Shida and Zhang, Kaicheng and Wang, Sen},
  journal={IEEE Transactions on Robotics},
  year={2025},
  publisher={IEEE}
}

If you use the Tank Dataset, please also cite:

@article{xu2025tank,
  title={Tank dataset: An underwater multi-sensor dataset for SLAM evaluation},
  author={Xu, Shida and Scharff Willners, Jonatan and Roe, Joshua and Katagiri, Sean and Luczynski, Tomasz and Petillot, Yvan and Wang, Sen},
  journal={The International Journal of Robotics Research},
  pages={02783649251364904},
  year={2025},
  publisher={SAGE Publications Sage UK: London, England}
}