A CLI tool for running DRL experiments on a simulated Autonomous vehicle platoon (or many).
To install, create a venv and install the requirements file. This program is loosely tested on Python 3.6, 3.7 and 3.8
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
Note: For simulations to render, you will need gym==0.21.0 pyglet==1.5.23
***Note that on windows you will perform venv/Scripts/Activate
instead of source venv/bin/activate
.
To see what the cli can do, enter the following.
python run.py --help
python run.py tr
To gain insight on configurable parameters, use
python run.py tr --help
A training session will create a .outputs folder within the base of the repo. This folder will contain an experiment folder containing the results.
After training, you can generate a latex report in a folder. This can be embedded into an overleaf project as a latex subfile if you like.
python run.py lmany
Note that this command only works if the .outputs folder contains an experiment(s).
If you wish to compare similar experiments, you can use at the accumulator.py file.
python run.py accumr
This will accumulate experiment results and plot the reward averaged across the seeds in svg files for each platoon.
You can re-run simulations on existing experiments, by running the below command and passing the file as an argument
python run.py esim
You can nest --help on any of the above commands if you cant figure out what to do with the cli.