HERO-SLAM: Hybrid Enhanced Robust Optimization of Neural SLAM

This is an official implementation of our work published in ICRA'24. Project Page

HERO-SLAM: Hybrid Enhanced Robust Optimization of Neural SLAM

Zhe Xin¹, Yufeng Yue², Liangjun Zhang³ and Chenming Wu³
¹Meituan UAV, Beijing, China, ²School of Automation, Beijing Institute of Technology, ³Robotics and Autonomous Driving Lab (RAL), Baidu Research

Paper (arXiv)

Installation

Please follow the instructions below to install the repo and dependencies.

git clone https://github.com/hero-slam/HERO-SLAM
cd HERO-SLAM

Install the environment

# Create conda environment
conda create -n heroslam python=3.7
conda activate heroslam

# Install the pytorch first (Please check the cuda version)
pip install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu113/torch_stable.html

pip install -r requirements.txt

# Build extension (marching cubes from neuralRGBD)
cd external/NumpyMarchingCubes
python setup.py install

Dataset

Please follow the instructions provided here: https://github.com/HengyiWang/Co-SLAM for downloading and preprocessing datasets.

Run

You can run HERO-SLAM using the code below.

For fixd interval, controlled by "ds_interval" in config file

python heroslam_opt_pose.py --config './configs/{Dataset}/{scene}.yaml

For adaptive interval, controlled by "run_interval" in config file, and "ds_interval = 1" in this mode.

python heroslam_opt_pose_adaptive_interval.py --config './configs/{Dataset}/{scene}.yaml

Evaluation

Code for evaluation strategy, performance analysis click here.

Acknowledgement

Our codebase builds on the code in Co-SLAM.

Citation

If you find our code or paper useful for your research, please consider citing:

@inproceedings{xin2024heroslam,
        title={HERO-SLAM: Hybrid Enhanced Robust Optimization of Neural SLAM},
        author={Zhe Xin, Yufeng Yue, Liangjun Zhang and Chenming Wu},
        booktitle={ICRA},
        year={2024}
}