Skip to content

Environments

Alvis Logins edited this page Apr 2, 2020 · 3 revisions

An environment is a problem with a minimal interface that an agent can interact with. The environments in the OpenAI Gym are designed in order to allow objective testing and bench-marking of an agents abilities.

Adding New Environments

  1. Write your environment in an existing collection or a new collection. All collections are subfolders of `/gym/envs'.
  2. Import your environment into the __init__.py file of the collection. This file will be located at /gym/envs/my_collection/__init__.py. Add from gym.envs.my_collection.my_awesome_env import MyEnv to this file.
  3. Register your env in /gym/envs/__init__.py:
register(
   	id='MyEnv-v0',
   	entry_point='gym.envs.my_collection:MyEnv',
)
  1. Add your environment to the scoreboard in /gym/scoreboard/__init__.py:
add_task(
   	id='MyEnv-v0',
   	summary="Super cool environment",
   	group='my_collection',
   	contributor='mygithubhandle',
)