Civilization Reinforcement Learning Agent

Inspiration

Civilization is cool

What it does

Dominates all other CPUs using Q-learning policies to decide whether to expand territory, make peace, stabilize, declare war, or sea invade. I also used Databricks Open Source Project, using the Delta Lake to store empire stats along with Mosaic to make cool visualizations of the trends of empires over time.

How we built it

Used a Civilization simulator https://github.com/ben9583/civilization-simulator which had a basic simulated Civilization game, and added an AI empire on top of it.

Challenges we ran into

It was hard to figure out how to add a separate empire to the impressive and established code. I found it difficult to figure out the action-value and state-value functions and how to implement the sequences of the deep learning algorithm.

Accomplishments that we're proud of

Creating the AI agent using the algorithm that could steamroll over the other empires.

What we learned

How to use the deep Q-learning algorithm

What's next for Civilization Reinforcement Learning Agent

Making it more efficient so my computer can cool down a bit

Built With

• databricks
• python