Skip to content

StanfordMSL/SousVide

Repository files navigation

SOUS VIDE

Installation

  1. Clone repository and load the submodules.
git clone https://github.com/StanfordMSL/SousVide.git
git submodule update --recursive --init
  1. Build ACADOS locally.
# Navigate to acados folder
cd <repository-path>/SousVide/FiGS/acados/

# Compile
mkdir -p build
cd build
cmake -DACADOS_WITH_QPOASES=ON ..
make install -j4

# Add acados paths to bashrc
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:"<acados_root>/lib"
export ACADOS_SOURCE_DIR="<acados_root>"
  1. Set up conda environment (in the main directory)
# Navigate to environment config location
cd <repository-path>/SousVide/

# Create and activate
conda env create -f environment_x86.yml
conda activate kitchen
  1. Download Example GSplats
# Navigate to gsplats parent folder
cd <repository-path>/SousVide/

# Download the zip-ed file below and unpack the contents (capture and workspace) into the gsplats folder
https://drive.google.com/file/d/1kW5dzsfD3rbRA3RIQDyJPG6_UJaO9ALP/view

Run SOUS VIDE Examples

Check out the notebook examples in the notebooks folder:

  1. figs_examples: Example code for generating GSplats and executing trajectories within them (using FiGS).
  2. sous_vide_examples: Use this notebook to try out two of the policies generated in the paper.

[COMING SOON: Oct 2025] Deploy SOUS VIDE in the Real World

Deploy SOUS VIDE policies on an MSL Drone. Tutorial and code coming soon!

About

Cooking Drone Visuomotor Policies in a Gaussian Splat Vacuum

Resources

License

Stars

Watchers

Forks

Packages

No packages published