A Reinforcement Learning Approach to Optimal Liquidity Provision in Uniswap v3

Team Members

Andy Hsu, Akshay Sreekumar

Description

This project simulates a decentralized liquidity pool (Uniswap v3) in Gym by formulating it as an MDP. We then use deep reinforcement learning to train liquidity provider agents in this environment. See report.pdf for more details.

Run Instructions

First, unzip the blockchain data to get the required CSV files: unzip blockchain_data.zip

Then simply run the training script: python train.py

You can edit the input parameters in config.yml