EffoNAV: an Effective Foundation-model-based Visual Navigation Approach for Challenging Environment.

EffoNAV can realize navigation under challenging situations such as large lighting variations and object variations.

conda env create -f train/train_environment.yml
conda activate EffoNAV
pip install -e train/

In this paper, we train on 4 publicly available datasets:

• RECON
• SCAND
• GoStanford
• SACSoN

You can use some sample scripts to process these datasets, either directly from a rosbag or from a custom format like HDF5s:

1. Run process_bags.py with the relevant args, or process_recon.py for processing RECON HDF5s.
2. Run data_split.py on your dataset folder with the relevant args.

Run this inside the ./train directory:

python train.py -c config/EffoNAV.yaml

You can use our pretrained checkpoints from here.

For the deployment procedures and details, please refer to Deployment

Our work references the training and deployment methods of visualnav-transformer. We are grateful for their contributions to the field of robot navigation.

@article{shen2025effonav,
  title={EffoNAV: An Effective Foundation-Model-Based Visual Navigation Approach in Challenging Environment},
  author={Shen, Wangtian and Gu, Pengfei and Qin, Haijian and Meng, Ziyang},
  journal={IEEE Robotics and Automation Letters},
  year={2025},
  publisher={IEEE}
}