NEURAL NETWORK TO PLAY CAR RACING GAME

Goals

• Implement a 2D Car Racing game environment using Pygame with possible play mode, autopilot mode and ai mode.
• Implement a Neural Network to play the game using Keras with categorical cross entropy loss function.

Libraries

• Pygame
• Pandas
• Numpy
• Keras

Tools and programming languages

• Python3
• Jupyter Notebook

Installation

• Make sure python3 and all libraries mentioned above are installed
• Clone this repo
• Navigate into the folder
• Open terminal and run

python3 main.py

Implementing your own Neural Network

Some of the parameters in main.py file are changeable. If you want to train your own Neural Network, open main.py and change some of the variables as shown below:

collect_data = True
ai = False
player_control = False
rows = 20000

In the code above collect_data = True means you can write all the state-action pairs into a csv file for future data analytics purposes. ai = False means you want to use an autopilot to play the game that will never fail, thus generate a proper state-action pairs to train you neural network player_control = False means no input from a player will be affecting a state except Esc button that exists the game rows = 20000 means after 20000 state-action pairs the csv file will be saved

Other changeable settings

You can go to constants.py file in constants folder, there You can set your own FRAME_RATE to influence how often the game frame will change in pygame, DATA_ROWS to limit amount of rows to be collected into a csv file, ACTION_PERFORM_RATE to decide how frequently your car will move ( Note: The new car will appear twice as rare as ACTION_PERFORM_RATE ) to make sure there is always a way out for your car during the game.

You can also change the environment by uploading your own images of cars into variables MY_CAR_ICON and ENEMY_CAR_ICON ( Note: for proper gaming experience make sure the image has sizes of 40px vs 80px )

You can change the background of the screen as well as the white vertical lines. BLACK = (0, 0, 0) stands for the color of background. Simply pick your favorite color and write its RGB code in the corresponding fields. GREY = (150, 150, 150) stands for vertical lines that separate the road lines