VizDoom-Reinforcement-Learning-Model

Reinforcement Learning / CNN model trained to play Doom within the VizDoom gym environment, for a variety of scenarios and complexities.

VizDoom Scenarios:

Basic

Trained to defeat a single unmoving enemy Can only move left, move right, and shoot ( more controls implemented later ) Reward based on if opponent was succesfully shot, and how fast/efficienct was this

Deadly Corridor

Trained to navigate a corridor with multiple attacking enemies Reward based on amount of damage taken and how fast level was completed

Defend the Center

Trained to defend the cenet of a room with enemies coming at all angles Reward based on how long the model can last in the center