Bomberman Bot
About
For the final project of Olin College's Software Design class, my team set out to write a bot
to play the classic arcade game, "Bomberman". We devised a way to train a neural net off a python emulation
of the game so that we could pull game data from the existing game and not rely on computer vision. This was also my
first experience making a GitHub pages site.
The full process of this project can be found on our documentation website
The project was resonably successful in terms of "surviving", i.e. moving around the map while placing bombs and avoiding enemies.
It was not great at clearing levels however as it could not reasonably predict enemies movements in relation to a bomb to eliminate the enemy.
Skills
- Python - The entirety of the bot was written in python
- Communication - Creating a website to fully document our process and presenting our work to the class
- Software Design - Generally thinking through the structures we built around the existing game to most efficently and easily train a neural net.
Map Kernal
A sliding box was held over the map to always keep the player character in the center. This made training the neural net computationally cheapere as it did not have to continuously find itself in the map each game tick.
The bot in action
Poster for final presentation