Inspiration
Inspired by Conway's Game of Life
What it does
Each ant robot in the game is controlled by a Neural Network. Ant robots are divided into different tribes. Ants from different tribes can fight against each other, and those from the same tribe can reproduce. A genetic algorithm is adopted to create the new-born ants.
How we built it
We use Echo State Network for the NN. Genes encode the readout matrix. Crossover by randomly swap the weights.
Challenges we ran into
The server does not support numpy and scipy. So the backend code cannot be run on the server.
Accomplishments that we're proud of
We came up with a good genetic algorithm for this agent control task.
What we learned
It's important to first install everything on the server.
What's next for Game of Evolution
Find a server that supports numpy and scipy, merge the backend and frontend, and set a website for it so that people can see how the world of ant agents evolve online.
Log in or sign up for Devpost to join the conversation.