- C 87.4%
- Python 8.3%
- HTML 1.9%
- Jupyter Notebook 0.7%
- Makefile 0.4%
- Other 1.1%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| DataS1 | ||
| real_world | ||
| sim | ||
| README.md | ||
Whisker-Based Tactile Flight for Tiny Drones
This repository contains the simulation, real-world implementation, and supplementary data/code release for the paper:
Whisker-Based Tactile Flight for Tiny Drones
Paper:
https://arxiv.org/abs/2510.03119
The goal of this project is to enable tiny drones to perform tactile navigation and exploration using whisker-like sensors and active exploration strategies.
Repository Overview
Whiskered-drone
│
├── simulation/
│ └── WhiskeredDroneSim/
│ ├── IsaacLab/
│ └── OmniDrones_whiskers/
│
├── real_world/
│ └── crazyflie-firmware/
│
└── DataS1/
├── [all data used in the paper]
└── whisker_repro_repo/
├── configs/
├── Dataset1
├── outputs
├── scripts/
├── src/
├── weights/
├── requirements.txt
└── README.md
Simulation
Simulation code is located in:
simulation/WhiskeredDroneSim/
It contains two main components:
IsaacLab
The base simulation framework built on top of Isaac Sim.
OmniDrones_whiskers
A modified drone simulation framework based on OmniDrones, integrated with Isaac Sim 4.1 and Isaac Lab, and extended with modules for whisker-based exploration.
Note that Isaac Sim 4.1 is not included in this repository and must be downloaded separately by users.
Detailed installation and setup instructions are provided in:
simulation/README.md
Real-world
The implementation used on the physical drone platform is located in:
real_world/
It includes:
- Modified Crazyflie firmware
- Whisker sensor interation
- Onboard tactile navigation and exploration
- Python scripts for running experiments
See:
real_world/README.md
DataS1: Supplementary data and reproduction code
The folder DataS1/ contains the supplementary material associated with the paper.
It includes:
- all data used in the paper
- reproduction code for the depth estimation experiments reported in the supplementary results
Reproduction code
The reproduction code is located in:
DataS1/whisker_repro_repo/
This sub-repository contains the code and pretrained weights used for reproducing the results in:
Table S1: Evaluation metrics for all models tested across trials in Dataset 1
Its contents include:
- classical baselines
- MLP training and testing
- full KF post-processing based on pretrained MLP models
- dataset configuration and evaluation scripts
- pretrained weights used in the released experiments
Please refer to:
DataS1/whisker_repro_repo/README.md
for detailed instructions on:
- dataset structure
- supported model families
- training and testing commands
- reproducing the full KF model
- required dependencies
Citation
If you use this code in your research, please cite:
@article{ye2025whisker,
title={Whisker-based Tactile Flight for Tiny Drones},
author={Ye, Chaoxiang and de Croon, Guido and Hamaza, Salua},
journal={arXiv preprint arXiv:2510.03119},
year={2025}
}
If you use the OmniDrones framework, please also cite their work.
Acknowledgement
This project builds upon the following open-source frameworks:
We thank the original authors for their contributions.
License
This repository follows the MIT License inherited from the OmniDrones project.
See the LICENSE file for details.