Flow Lite (Name TBD still)

Installation

1. Make sure that your computer's or server's OS is Ubuntu 18.04 or lower, or Mac.
2. Follow instructions here to install Miniconda, likely wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh followed by bash Miniconda3-latest-Linux-x86_64.sh
3. Run bash scripts/setup_sumo_<os_version>.sh corresponding to your OS version to set up SUMO and add ~/sumo_binaries/bin to your PATH environment variable. Try running sumo
4. Install PyTorch from pytorch.org.
5. Clone the util directory with git clone [email protected]:ZhongxiaYan/util.git u or git clone https://github.com/ZhongxiaYan/util.git u and pip install -r u/requirements.txt
6. Install dependencies pip install -r requirements.txt

Run Instructions

Let <res_dir> be the result directory, which is where the model checkpoints, training logs, and training csv results will be saved. Add render as an argument for using sumo-gui instead of sumo. E.g. python pexps/<script>.py <res_dir> render.

Bottleneck

python pexps/bneck.py <res_dir>

Intersection

python pexps/g2x1.py <res_dir>

Intersection (8-way)

This is currently incomplete and only has some SUMO support for custom 8-way networks, but RL support is not implemented at the moment. python pexps/8way.py <res_dir>