Notes

experiment.py for room 0 experiments

experimentOffice3.py for office 3 experiments

experimentOffice4.py for office 4 epxeriments

Use all respective .npy and .pth files for the Robot class, included within the code and the repo.

Enforcing safety for vision-based controllers via Control Barrier Functions and Neural Radiance Fields

Installation

All the dependencies should be ensured in place. Please use anaconda to create an environment called nice-slam. For linux, libopenexr-dev is required before creating the environment.

sudo apt-get install libopenexr-dev
    
conda env create -f environment.yaml
conda activate nice-slam

Training

Firstly you need to download the Replica data as below, which will be saved into ./Datasets/Replica folder.

bash scripts/download_replica.sh

Then you can train NICE-SLAM:

python -W ignore run.py configs/Replica/room1.yaml

The mesh file is saved as $OUTPUT_FOLDER/mesh/final_mesh_eval_rec.ply, where the unseen regions are removed using all the frames.

For more datasets, please view the instruction of NICE-SLAM.

Experiments

To observe the performance of the controller in single-integrator systems, you need to run:

python single_cbf.py configs/Replica/room1.yaml --output output/Replica/room1

and for double-integrator systems:

python double_cbf.py configs/Replica/room1.yaml --output output/Replica/room1

The visualized results are automatically saved as video.avi.

Citation

@article{tong2022enforcing,
  title={Enforcing safety for vision-based controllers via Control Barrier Functions and Neural Radiance Fields},
  author={Tong, Mukun and Dawson, Charles and Fan, Chuchu},
  journal={arXiv preprint arXiv:2209.12266},
  year={2022}
}