Skip to content

CorentinVermeulen/epl_flappybird

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 

Repository files navigation

EPL FlappyBird

This repository contains the code for a master thesis at UCLouvain. The aim of the project was to train a DQN agent to play FlappyBird and to explore how various environmental modifications impact the training process. Among these modifications, we include:

  • Training on fixed or random pipes
  • Training with some randomness in the Jump Force
  • Training with increased action space

The code is based on the following sources:

The code is structured as follows in the source directory:

  • flappy_bird_gym: contains the FlappyBird game and the environment modifications
  • run_baseline_example: Example of the output of a training
  • agent_simple, agent_multi, agent_RGB: DQN agents with some modifications (simple is the classic agent, RGB is the agent for RGB images, multi is the agent with increased action space)
  • evaluation: code to evaluate the results of the training and create the plots
  • exp_X: code to run the experiments
  • exp_X_test: code to test the experiments
  • utils: some utility functions

Releases

No releases published

Packages

No packages published

Languages