Custom MuJoCo Environment in Openai Gym
Pre-Requisites
MuJuCo is a proprietary software which can be used for physics based simulation. First thing is to get a license as described in here. Then install mujoco-py as described in the Readme.md.
Procedure
The general structure of the package creation for registering openai-gym environments is as follows
FolderA/
README.md
setup.py
FolderB/
__init__.py
envs/
__init__.py
MyRobotEnv.py
README.md usually contain description of the project. The FolderA/setup.py should contain the following lines
from setuptools import setup
setup(name='FolderB',
version='0.1',
install_requires=['gym',
'numpy']) # And any other dependencies required
The FolderA/FolderB/__init__.py should have
from gym.envs.registration import register
register(id='RobotName-v0',
entry_point='FolderB.envs:EnvClassName',)
EnvClassName is the class described inside MyRobotEnv.py file. The file FolderA/FolderB/envs/__init__.py should contain
from FolderB.envs.MyRobotEnv import EnvClassName
Setup MuJoCo Environment
The existing MuJoCo environments are available here. MuJoCo XML files for several robots and an extensive documentations are published here(openai-github, Modeling). The structure of the MyRobotEnv.py will be as follows.
import numpy as np
import os
from gym import utils, error, spaces
from gym.envs.mujoco import mujoco_env
from mujoco_py import MjViewer, functions
class EnvClassName(mujoco_env.MujocoEnv, utils.EzPickle):
def __init__(self):
utils.EzPickle.__init__(self)
FILE_PATH = '' # Absolute path to your .xml MuJoCo scene file
# OR.
# FILE_PATH = os.path.join(os.path.dirname(__file__), "MyRobot.xml")
frame_skip = 5
mujoco_env.MujocoEnv.__init__(self, FILE_PATH, frame_skip)
def step(self, a):
# Carry out one step
# Don't forget to do self.do_simulation(a, self.frame_skip)
def viewer_setup(self):
# Position the camera
def reset_model(self):
# Reset model to original state.
# This is called in the overall env.reset method
# do not call this method directly.
def _get_obs(self):
# Observation of environment fed to agent. This should never be called
# directly but should be returned through reset_model and step
Installation
- Inside the folder where the
setup.pyis, run
pip install -e .
Test Installation
import gym
import FolderB
env = gym.make('RobotName-v0')
Enjoy Reading This Article?
Here are some more articles you might like to read next: