Checkout environment.md for details.
Download Simstar from the link:
https://drive.google.com/open?id=1Gp-XXnOX9dbDcfqFJNJ4UtZqo9sWqjUg
Download V.2.01 from the link above. Please reinstall simstar PythonAPI for the update to be compatible.
Change Log
- New tracks are added: Circular and Straight
- Flag marks are added ever 200m of the tracks
- evaluate.py is updated to examplify a race with trained agents
Just click on Simstar.exe and Simstar is ready.
Install Vulkan Libraries
sudo apt-get install vulkan-utils
cd Simstar
chmod 777 -R *
./Simstar.sh
Create a new environment using anaconda.
conda env create --file environment.yml
conda activate simstar-env
Install required python libraries from requirements.txt by
pip install -r requirements.txt
The final evaluation will be using pytorch version 1.5 and CUDA version 10.2.
cd PythonAPI
python setup.py install --user
There are multiple stages that needs to be checked.
Open the simstar executable, allow for networking if asked.
cd sac_example
python example_experiment.py
To test a closed loop training with Pytorch, you can run the example SAC agent from examples folder.
cd sac_example
python train_example.py