Getting Started with MuJoCo

MuJoCo

Half Cheetah in MuJoCo

MuJoCo stands for Multi-Joint dynamics with Contact. It is a general purpose physics engine that aims to facilitate research and development in robotics, biomechanics, graphics and animation, machine learning, and other areas that demand fast and accurate simulation of articulated structures interacting with their environment. Initially developed by Roboti LLC, it was acquired and made freely available by DeepMind in October 2021, with the goal of making MuJoCo an open-source project.

MuJoCo is a physics engine that aims to facilitate research and development in robotics, biomechanics, graphics and animation, and other areas where fast and accurate simulation is needed.

MuJoCo offers a unique combination of speed, accuracy and modeling power, yet it is not merely a better simulator. Instead it is the first full-featured simulator designed from the ground up for the purpose of model-based optimization, and in particular optimization through contacts.

MuJoCo makes it possible to scale up computationally-intensive techniques such optimal control, physically-consistent state estimation, system identification and automated mechanism design, and apply them to complex dynamical systems in contact-rich behaviors. It also has more traditional applications such as testing and validation of control schemes before deployment on physical robots, interactive scientific visualization, virtual environments, animation and gaming.

The Documentation for MuJoCo is available here

Installation

Let us now follow the following steps to setup MuJoCo on our Linux System

1. Install anaconda Download

sudo chmod +x Anaconda3-2021.11-Linux-x86_64.sh

./Anaconda3-2021.11-Linux-x86_64.sh

2. Install git sudo apt install git

3. install the mujoco library

• Download the Mujoco library from here
• create a hidden folder :

mkdir /home/username/.mujoco

• extract the library to the .mujoco folder
• include these lines in .bashrc file:

export LD_LIBRARY_PATH=/home/tayal/.mujoco/mujoco210/bin

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/nvidia

export PATH="$LD_LIBRARY_PATH:$PATH"

export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libGLEW.so

• source .bashrc
• Test that the library is installed by going into:

cd ~/.mujoco/mujoco210/bin

./simulate ../model/humanoid.xml

4. Install mujoco-py:

conda create --name mujoco_py python=3.8

conda activate mujoco_py

sudo apt update

sudo apt-get install patchelf

sudo apt-get install python3-dev build-essential libssl-dev libffi-dev libxml2-dev

sudo apt-get install libxslt1-dev zlib1g-dev libglew1.5 libglew-dev python3-pip

git clone https://github.com/openai/mujoco-py cd mujoco-py

pip install -r requirements.txt

pip install -r requirements.dev.txt

pip3 install -e . --no-cache

5. Reboot your machine

6. Run these commands

conda activate mujoco_py

sudo apt install libosmesa6-dev libgl1-mesa-glx libglfw3 sudo ln -s /usr/lib/x86_64-linux-gnu/libGL.so.1 /usr/lib/x86_64-linux-gnu/libGL.so

pip3 install -U 'mujoco-py<2.2,>=2.1'

cd examples python3 setting_state.py

MuJoCo Tutorial

You can go through the following tutorial to get a hang of MuJoCo Environment:

MuJoCo Tutorial Github Link

The main purpose of this repo is providing the starter code required to run a MuJoCo simulation with keyboard and mouse callbacks using its Python bindings. The base class is in mujoco_base.py. To create your own MuJoCo simulation, you can create a new class that inherits mujoco_base.MuJoCoBase. An example of this usage is provided in example_projectile.py, the new class should implement the functions.

- reset()       # Initializes the enviroment and control callback
- controller()  # Adds control actions
- simulate()    # Copy the simulate() function from 
                # mujoco_base.MuJoCoBase and add your own twist

MuJoCo Bootcamp Examples

- Projectile with drag
- Control a simple pendulum
- Control a double pendulum
- Leg swing
- Manipulator drawing
- Control an underactuated pendulum
- Gymnast swing/release on a bar
- 2D Hopper
- Initial Value Problem
- Inverse Kinematics
- 2D Biped