/TURTLMap

Textureless Underwater Real Time Localization and Mapping

Primary LanguageC++

šŸ¢TURTLMap: Textureless Underwater Real-Time Localization and Mapping

We propose the first real-time localization and dense mapping solution that is robust to low-texture underwater scenes. TURTLMap is designed to be low-cost for scalability and is capable of running onboard on an NVIDIA Jetson Orin Nano in real-time. TURTLMap leverages GTSAM to develop a multi-modal robust odometry and Voxblox to build a dense 3D map.

TURTLMap in action

Update

  • (2024/09): We have officially released code and data for TURTLMap!
  • (2024/06): šŸ”„TURTLMap is getting a lot of attention on LinkedIn!
  • (2024/06): TURTLMap is accepted to IROS 2024! See you in Abu Dhabi!

Installation

Please refer to the installation guide for detailed instructions.

Data Preparation

We provide logs collected from our field experiments. Please refer to the data preparation guide for detailed instructions on downloading and using them to run TURTLMap.

Use TURTLMap

For instructions on how to run, visualize, and evaluate TURTLMap, please refer to the user guide.

TURTLMap on Your Own Robot

For instructions on how to extend TURTLMap to your own robot, please refer to the extension guide.

Citation

Please cite our paper if you find TURTLMap useful for your research:

@article{song2024turtlmap,
  title={TURTLMap: Real-time Localization and Dense Mapping of Low-texture Underwater Environments with a Low-cost Unmanned Underwater Vehicle},
  author={Song, Jingyu and Bagoren, Onur and Andigani, Razan and Sethuraman, Advaith Venkatramanan and Skinner, Katherine},
  journal={arXiv preprint arXiv:2408.01569},
  year={2024}
}