Wisaacj/Autonomous-Lane-Switching

2D autonomous lane switching using deep reinforcement learning

Autonomous Control in a Highway Environment with DRL

You can find the highway environment here.

Rainbow DQN

We implemented following algorithms based off of Deep Q-Networks:

• DQN - See also here and here
• Double DQN
• DQN with Prioritised Experience Replay
• Dueling DQN
• DQN with Noisy Networks for Exploration
• Categorical DQN
• DQN with N-step Learning
• Rainbow DQN

We used this tutorial to help us implement the above algorithms.

PPO

You can find our implementation of PPO here. We partially based it paritially on Eric Yang Yu's Medium Guide to implementing PPO

To run this code

Create a virtual environment:

python3 -m venv venv

Activate your virtual environment (Unix):

source venv/bin/activate

Or, activate your virtual environment (Windows):

.\venv\bin\activate

Install the required packages:

pip install -r requirements.txt

And you should be set to run each of the notebooks! (as long as you've been blessed by the Cuda gods)