SOUS VIDE

Installation

1. Clone repository and load the submodules.

git clone https://github.com/StanfordMSL/SousVide.git
git submodule update --recursive --init

2. 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>"

3. 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 sv-env

4. Download Example GSplats

# Navigate to gsplats parent folder
cd <repository-path>/SousVide/

# Use gdown to download
gdown --folder https://drive.google.com/drive/folders/1Q3Jxt08MUev_jWzHjpdltze7X4VArsvA?usp=drive_link --remaining-ok

# Alternatively, you can 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. sv_shakedown: Use this notebook to verify all components before running other SV notebooks. It does not produce a usable policy
3. sv_robustness: Produces the policies used in Section VI.A and VI.B.
4. sv_extended: Produces the policy for the extended trajectory in Section VI.C.
5. sv_cluttered: Produces the policy for the cluttered environment trajectory in Section VI.C.

[COMING SOON: 04/25] Deploy SOUS VIDE in the Real World

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