Lunar Lander - CT213 Exam

This project uses the Lunar Lander environment from OpenAI's Gym library.

Group:

Lucas Ribeiro do Rêgo Barros (2°FUND - leader)

Luiz Felipe Vezzali de Paula (1° AER)

Mateus Pereira Alves (1° AESP)

GitHub: https://github.com/rbarroslucas/ct213_exam

Project goal

This project aims to solve the Lunar Lander environment with DQN, Double DQN and Dueling DQN algorithms. The project's paper can be found between the files.

Folder Results

The folder Results presents some trained agents with the raw results

Environment Setup

To set up the Conda environment, follow these steps:

1. Create the Conda environment using the existing one using anaconda navigatir by one of the two .yaml files (prefer the ct_exam.yaml) using Python 3.10.9

2. Activate the environment:

   conda activate exam_env

obs: The CT213 environment (from classroom) was observed to work with all dependencies in the project

Running the Project

After activating the Conda environment, you can run the main code that :

python src/main.py

Testing different agents

You can test different agents by choosing the configurations on src/utils/params.py

Evaluating a pre-trained agent

If you don't wanna train a agent and just evaluate, you can run the script src/only_evaluation.py