Skip to content

Latest commit

 

History

History
8 lines (8 loc) · 590 Bytes

File metadata and controls

8 lines (8 loc) · 590 Bytes

DOOM&RL - ViZDoom + DQN

The scope of the project is to develop a DQN agent, capable of playing different scenarios of the game DOOM. The computed models are able to complete the basic scenario and the defend the center level with satisfactory results.

How to run it?

Simply runs the main files, each of them is associated to a scenario and a technique, inside each of them there's a flag called case, switch it between 'training' and 'testing' based on what you want.

Final results - BASIC

Final results - DEFEND THE CENTER