The aMAZEing Race

Inspiration

Our inspiration is the classic idea of how a rat runs mazes, and we wanted to create something similar to that purely through machine learning

What it does

It takes control over a player and tries its very best to reach a target in a maze in a timely manner.

How we built it

We wrote the scripts for this program with the help of Tensorflow and Unity. Through a learning script the bot has trained itself progressively using Reinforcement Learning and over time has gotten better and better at completing its task.

Challenges we ran into

The biggest challenge that we ran into was how to properly adjust the weights to have the bot train as time efficiently as possible. With some algorithms it would struggle to even get up while in others it would sometimes make it to the end of the maze but not consistently. None of us are experts when it comes to using the Unity Engine so we had to go through a bit of a struggle to get certain collisions to be detected.

Accomplishments that we're proud of

We're very proud of with some of our algorithms it managed to consistently make its way towards the target in some of the mazes.

What we learned

We learned the basics of implementing machine learning. We had researched the concepts behind machine learning but this is our first time actually implementing them into an actual project.

What's next for The aMAZEing Race

We hope to merge the functionality of the two neural networks we created and hopefully expand this project in the future, possibly adding the concept of jumping and playing around with other models and mazes.