QRSAC
Implementation of Quantile Regression Soft Actor Critic (QRSAC) from “Outracing champion Gran Turismo drivers with deep reinforcement learning” by Peter R. Wurman, Samuel Barrett, Kenta Kawamoto, James MacGlashan, Kaushik Subramanian, Thomas J. Walsh, Roberto Capobianco, Alisa Devlic, Franziska Eckert, Florian Fuchs, Leilani Gilpin, Piyush Khandelwal, Varun Kompella, HaoChih Lin, Patrick MacAlpine, Declan Oller, Takuma Seno, Craig Sherstan, Michael D. Thomure, Houmehr Aghabozorgi, Leon Barrett, Rory Douglas, Dion Whitehead, Peter Dürr, Peter Stone, Michael Spranger & Hiroaki Kitano. [Paper].
This repository is based on RLkit and DSAC, two popular reinforcement learning frameworks implemented in PyTorch.
The github code link: [Code]
Requirements
- python 3.10+
- All dependencies are available in requirements.txt and environment.yml
Usage
You can write your experiment settings in configs/your_config.yaml and run with
python qrsac.py --config your_config.yaml --gpu 0 --seed 0
Set --gpu -1, your program will run on CPU.
Experiments
2 different experiments are conducted to validate the working of the QRSAC algorithm - on a donkercar simulator and on a real-world scaled RC car (Jetracer).
Experiments on DonkeyCar
Experiments on JetRacer
