/husky_nav

Navigation pipelines with Traversability Estimation tested with real Clearpath Husky robot and in Gazebo simulator

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Husky MonoForce

Navigation with Husky robot only from RGB images. The ctu-vras/monoforce package is used as robot-terrain interaction and path planning pipeline.

5 minute CVPR presentation video link

Trajectories prediction is based on the NVIDIA/warp and ctu-vras/diffsim packages.

Husky in Gazebo Husky in Gazebo

Navigation consists of the following stages:

  • Image acquisition: RGB images are acquired from the Gazebo simulator.
  • Height map prediction: The Terrain Encoder part of the MonoForce is used to estimate terrain properties.
  • Trajectories prediction: The Diff Physics part of the MonoForce is used to shoot the robot trajectories.
  • Trajectory selection: The trajectory with the smallest cost based on robot-terrain interaction forces is selected.
  • Control: The robot is controlled to follow the selected trajectory.

Installation

Please, refer to INSTALL.md for installation instructions.

Usage

roslaunch husky_nav husky_monoforce.launch warp_dphys:=true

Citation

Consider citing the paper if you find the package relevant to your research:

@article{agishev2023monoforce,
    title={MonoForce: Self-supervised Learning of Physics-aware Model for Predicting Robot-terrain Interaction},
    author={Ruslan Agishev and Karel Zimmermann and Vladimír Kubelka and Martin Pecka and Tomáš Svoboda},
    year={2023},
    eprint={2309.09007},
    archivePrefix={arXiv},
    primaryClass={cs.RO}
}