An evolutionary neural network trained to play a clone of the original NES Super Mario Bros.
This project contains the source code for an evolutionary algorithm optimized artificial neural network trained to play a clone of the first level of Super Mario Bros.
All code was written in Java using the libGDX game development framework and TensorFlow machine intelligence library.
- Install Java JDK
- Install Gradle
- Clone this repository
git clone https://github.com/artur-sak13/MarioFlowGDX.git
- Open the project in a Java supporting IDE (i.e. Eclipse, IntelliJ, Netbeans, etc.)
- Import the Gradle project settings
- Edit your run configuration and make sure that
core/assets
is set as your working directory - Run the project and have fun!
You can either run com.artursak.mariobros.desktop as a Java application to build the project or run the .jar
file
java -jar MarioBros.jar
- Left arrow or A -> Move Left
- Right arrow or D -> Move Right
- Up arrow or W -> Jump
- Spacebar -> Fireball
Note: Fire Mario and stars items are not yet implemented so Mario can throw fireballs from the start.
Check out all of my projects!