RL

RL solving Highway environments

The goal of this project is to create differents agents to sole differents highway environments with discrete sets of actions or states.

Group

• Erwan DAVID
• Guillaume FAYNOT

Run the simulation

Train your agent on a specific environment by one of the following command in the appropriate folder :

python3 main1.py python3 main2.py python3 main3.py

Run the appropriate notebook (main1.ipynb, main2.ipynb or main3.ipynb) to try your agent in the environment.

Models

The project includes three different environments solved with various RL algorithms:

1. Environment 1: A classic highway environment solved using DQN with continuous states and discrete actions.
2. Environment 2: A racetrack environment solved using PPO with discrete states and continuous actions.
3. Environment 3: A highway merging scenario solved using Stable-Baselines3.

For more details, refer to the corresponding reports available in the repository.