AIvysuar (Pokemon Showdown RL Agent)

Inspiration

The inspiration for this project came from our collective love for Pokemon and reinforcement learning models!

What it does

AIvysuar is a reinforcement learning agent designed to play Pokemon Showdown. It learns optimal battle strategies through automated battles, reading the chat, and adapting to its opponent's strategies.

How we built it

We used Python, PyTorch, Pokemon Showdown API, and the poke-env library. We used Deep Q-Networks to train our reinforcement learning AI agent.

Challenges we ran into

We ran into issues like high dimensional space complexity due to the sheer complexity of Pokemon Showdown as well as overfitting to specific strategies.

Accomplishments that we're proud of

• Creating a function RL agent that can play Pokemon Showdown at a competitive level
• Successfully implementing a Deep Q-Network
• The agent has beaten multiple real players
• Chatbot that reacts to player's moves

What we learned

As a team we gained experience in reinforcement learning and game theory.

What's next for AIvysaur

• Expand AIvysaurs adaptability to a wider range of strategies
• Create more agents to train against AIvysaur
• Optimize AIvysaurs performance
• Placing in the top 100 of the Pokemon Showdown competitive ladder

Built With

• api
• poke-env
• python
• pytorch
• q-networks