Doom Reinforcement Learning

This project explores using reinforcement learning to train an agent to play Doom

Single-Threaded Version

The single-threaded implementation achieved the following results:

• At 100k iterations: It can beat level 1.
• At 250k iterations: It can beat level 2.
• At 1 million iterations: It can beat level 1 and then level 2 sequentially.

Performance and Runtime

Training was conducted on my desktop with an RTX 3070 Ti. The training time was approximately 1 hour per 500k iterations.

"Multi-Threaded" Implementation

TODO: Implement and document the multi-threaded version for improved performance.

How to Run This Yourself

Frankly, I doubt anyone cares enough to do so, but I will include this nonetheless.

Requirements

• Python: 3.6.0 - 3.9.0
• Stable-Baselines3: 1.5.0
• Pip: 21.0 (or newer)
• Gym: 0.21.0
• Wheel: 0.38.0
• VizDoom: 1.2.3
• Setuptools: 65.5.0

(All packages can be installed via pip)

pip install stable-baselines3==1.5.0 gym==0.21.0 vizdoom==1.2.3 wheel==0.38.0 setuptools==65.5.0