vladimirsvsv77/flappy-bird-gymnasium

flappy-bird-gymnasium with Jupyter notebooks of RL algorithms.

Flappy Bird for Gymnasium

This repository contains the implementation of several reinforcent learning algorithms for the Flappy Bird game. The implementation of the game's logic and graphics was based on the flappy-bird-gym project, by @Talendar. The gymnasium fork of the original project is the basis of this project, as implemented by @markub3327.

State space

The "FlappyBird-v0" environment, yields simple numerical information about the game's state as observations representing the game's screen.

FlappyBird-v0

There exist two options for the observations:

1. option

• The LIDAR sensor 180 readings (Paper: Playing Flappy Bird Based on Motion Recognition Using a Transformer Model and LIDAR Sensor)

2. option

• the last pipe's horizontal position
• the last top pipe's vertical position
• the last bottom pipe's vertical position
• the next pipe's horizontal position
• the next top pipe's vertical position
• the next bottom pipe's vertical position
• the next next pipe's horizontal position
• the next next top pipe's vertical position
• the next next bottom pipe's vertical position
• player's vertical position
• player's vertical velocity
• player's rotation

Action space

• 0 - do nothing
• 1 - flap

Rewards

• +0.1 - every frame it stays alive
• +1.0 - successfully passing a pipe
• -1.0 - dying
• −0.5 - touch the top of the screen

Installation

To install flappy-bird-gymnasium, simply run the following command:

$ pip install flappy-bird-gymnasium

Usage

Like with other gymnasium environments, it's very easy to use flappy-bird-gymnasium. Simply import the package and create the environment with the make function. Take a look at the sample code below:

import flappy_bird_gymnasium
import gymnasium
env = gymnasium.make("FlappyBird-v0", render_mode="human", use_lidar=True)

obs, _ = env.reset()
while True:
    # Next action:
    # (feed the observation to your agent here)
    action = env.action_space.sample()

    # Processing:
    obs, reward, terminated, _, info = env.step(action)
    
    # Checking if the player is still alive
    if terminated:
        break

env.close()

Playing

To play the game (human mode), run the following command:

$ flappy_bird_gymnasium

To see a random agent playing, add an argument to the command:

$ flappy_bird_gymnasium --mode random

To see a Deep Q Network agent playing, add an argument to the command:

$ flappy_bird_gymnasium --mode dqn