Kuan Xu1, Yuefan Hao2, Shenghai Yuan1, Chen Wang2, Lihua Xie1
1: Centre for Advanced Robotics Technology Innovation (CARTIN), Nanyang Technological University
2: Spatial AI & Robotics Lab (SAIR Lab), Department of Computer Science and Engineering, University at Buffal
📄 [Arxiv] | 💾 [Project Site] |
AirSLAM is an efficient visual SLAM system designed to tackle both short-term and long-term illumination challenges. Our system adopts a hybrid approach that combines deep learning techniques for feature detection and matching with traditional backend optimization methods. Specifically, we proposea unified convolutional neural network (CNN) that simultaneously extracts keypoints and structural lines. These features are then associated, matched, triangulated, and optimized in a coupled manner. Additionally, we introduce a lightweight relocalization pipeline that reuses the built map, where keypoints, lines, anda structure graph are used to match the query frame with themap. To enhance the applicability of the proposed system to real-world robots, we deploy and accelerate the feature detection and matching networks using C++ and NVIDIA TensorRT. Extensive experiments conducted on various datasets demonstrate that our system outperforms other state-of-the-art visual SLAM systems in illumination-challenging environments. Efficiency evaluations show that our system can run at a rate of 73Hz on a PC and 40Hz on an embedded platform.
- [2024.08] We release the code and paper for AirSLAM.
- [2023.07] AriVO is accepted by IROS 2023.
- [2022.10] We release the code and paper for AirVO. The code for AirVO can now be found here.
- OpenCV 4.2
- Eigen 3
- Ceres 2.0.0
- G2O (tag:20230223_git)
- TensorRT 8.6
- CUDA 12
- python
- ROS noetic
- Boost
docker pull xukuanhit/air_slam:v4
docker run -it --env DISPLAY=$DISPLAY --volume /tmp/.X11-unix:/tmp/.X11-unix --privileged --runtime nvidia --gpus all --volume ${PWD}:/workspace --workdir /workspace --name air_slam xukuanhit/air_slam:v4 /bin/bashThe data for mapping should be organized in the following Automous Systems Lab (ASL) dataset format:
dataroot
├── cam0
│ └── data
│ ├── 00001.jpg
│ ├── 00002.jpg
│ ├── 00003.jpg
│ └── ......
├── cam1
│ └── data
│ ├── 00001.jpg
│ ├── 00002.jpg
│ ├── 00003.jpg
│ └── ......
└── imu0
└── data.csv
After the map is built, the relocalization requires only moocular images. Therefore, you only need to place the query images in a folder.
cd ~/catkin_ws/src
git clone https://github.com/sair-lab/AirSLAM.git
cd ../
catkin_make
source ~/catkin_ws/devel/setup.bash
The launch files for VO/VIO, map optimization and relocalization are placed in VO folder, MR folder, and Reloc folder, respectively. Before running them, you need to modify the corresponding configurations according to you data path and the desired map saving path. The following is an example of mapping, optimization, and relocalization with the EuRoC dataset.
1: Change "dataroot" in VO launch file to your own data path. For the EuRoC dataset, "mav0" needs to be included in the path.
2: Change "saving_dir" in the same file to the path where you want to save the map and trajectory. It must be an existing folder.
3: Run the launch file:
roslaunch air_slam vo_euroc.launch
1: Change "map_root" in MR launch file to your own map path.
2: Run the launch file:
roslaunch air_slam mr_euroc.launch
1: Change "dataroot" in Reloc launch file to your own query data path.
2: Change "map_root" in the same file to your own map path.
3: Run the launch file:
roslaunch air_slam reloc_euroc.launch
Launch folder and config folder respectively provide the lauch files and configuration files for other datatsetsin the paper. If you want to run AirSLAM with your own dataset, you need to create your own camera file, configuration file, and launch file.
- Initial release. 🚀
- Support SuperGlue as feature matecher
- Optimize the TensorRT acceleration of PLNet
@article{xu2024airslam,
title = {{AirSLAM}: An Efficient and Illumination-Robust Point-Line Visual SLAM System},
author = {Xu, Kuan and Hao, Yuefan and Yuan, Shenghai and Wang, Chen and Xie, Lihua},
journal = {arXiv preprint arXiv:2408.03520},
year = {2024},
url = {https://arxiv.org/abs/2408.03520},
code = {https://github.com/sair-lab/AirSLAM},
}
@inproceedings{xu2023airvo,
title = {{AirVO}: An Illumination-Robust Point-Line Visual Odometry},
author = {Xu, Kuan and Hao, Yuefan and Yuan, Shenghai and Wang, Chen and Xie, Lihua},
booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
year = {2023},
url = {https://arxiv.org/abs/2212.07595},
code = {https://github.com/sair-lab/AirVO},
video = {https://youtu.be/YfOCLll_PfU},
addendum = {SAIR Lab Recommended}
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent