Summary

In this project, our goal is to let agent protect villagers and kill zombies. We do this by using deep reinforcement learning, which is a combination of deep learning and reinforcement learning. By randomly choosing actions at the beginning, agent will learn from the algorithm and start to choose the best action or move it should do to accomplish our goal.

Video Introduction

Click Me

Code

Source code

Reports

• Proposal
• Status
• Final

Useful Resources

Some of the link maynot be relevant to Malmo, but we are using similar idea here.

Fighting Zombies in Minecraft With Deep Reinforcement Learning - Hiroto Udagawa, Tarun Narasimhan, Shim-Young Lee

Find path using Q-learning

Beat Atari with Deep Reinforcement Learning! - Adrien Lucas Ecoffet

Improvements in Deep Q Learning: Dueling Double DQN, Prioritized Experience Replay, and fixed Q-targets - Thomas Simonini