Skip to content

Latest commit

 

History

History
19 lines (16 loc) · 2.52 KB

README.md

File metadata and controls

19 lines (16 loc) · 2.52 KB
Raw

Classic Control

We wrap the classic controls problems first introduced by Richard Sutton into the domain of RL. These are ported as is from OpenAI's GYM. A nice hello world on-boarding to cherry can be performed by solving CartPole-v0 with VPG outlined here.

Arcade Learning Environment (ALE)

We wrap the Arcade Learning Environment (ALE) which includes some of the classic Atari 2600 games.. We further wrap the GYM env to include few prosed changes to default environment from Deepmind outlined here. These changes include

VizDoom

Doom is classic FPS game form the the early 90s. Is available both in single and multi-agent environment. The authors provided an easy to use pythonic interface for RL community to use. We wrap vizdoom with minimal changes. We include the few scenarios which are popular with RL community to try. Including the Hello-worldish basic scenario.

  • Basic : Single target at the back of the wall
  • Death Match : Multi agent death match
  • Health Gathering : Learn to survive by gather med-packs
  • Deadly Corridor : Learn to navigate a maze & survive by terminating zombies

PyBullet

PyBullet provides a convenient non-commercial equivalent to Mujoco. This environment includes most of the environments included in Mujoco and more. OpenAI's gym includes a succinct description their support for PyBullet.