akolishchak / doom-net-pytorch

Reinforcement learning models in ViZDoom environment

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

DoomNet

DoomNet is a ViZDoom agent trained by reinforcement learning. The agent is a neural network that outputs a probability of actions given only pixels from the screen buffer and set of game variables.
DoomNet is a 1st Runner-Up at Visual Doom AI Competition 2018.

DoomNet on a simple behavior tree

DoomNet track1, submission 0

Joint work with Bobby DeSimone

DoomNet's view is at left in the middle row
DoomNet track1, elimination round 2017

D3 Battle

DoomNet trained on D3-battle config

Health Gathering

DoomNet trained on health gathering config

Rocket Basic

DoomNet trained on basic rocket config

Basic

DoomNet trained on basic config

About

Reinforcement learning models in ViZDoom environment

License:MIT License


Languages

Language:Python 88.7%Language:Shell 11.3%