BolunDai0216 / PyMuJoCoBase

PyMuJoCoBase

This repo contains starter code and examples for running simulations in MuJoCo with its Python bindings. This repo is inspired by the C code developed for Prof.Pranav Bhounsule's MuJoCo Bootcamp.

Installation

First, install the Python bindings for MuJoCo (newly tested on v2.3.0)

python3 -m pip install mujoco

Then, run the following commands to install PyMuJoCoBase

git clone https://github.com/BolunDai0216/PyMuJoCoBase.git
cd PyMuJoCoBase
python3 -m pip install .

Contents

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

How to apply control

The sole purpose of the controller() function is to change the values of the data.ctrl which corresponds to the actuator control values. There are two ways to achieve this, one is to set controller() as the control callback function

mj.set_mjcb_control(self.controller)

in reset(), an example usage is shown in example_manipulator_drawing.py. This tells MuJoCo to run the controller automatically. Another way to is to run controller() directly in simulate(), i.e., add

while (self.data.time - simstart < 1.0/60.0):
    self.controller(self.model, self.data)
    mj.mj_step(self.model, self.data)

to simulate(), please refer to example_manipulator_ik.py for this usage. This is a more manual approach, which applies the new control action at each time step.

MuJoCo Bootcamp Examples

All of the examples in the MuJoCo Bootcamp are translated into Python. The examples include:

- Projectile with drag (pymjbase-projectile-example)
- Control a simple pendulum (pymjbase-pendulum-example)
- Control a double pendulum (pymjbase-dbpendulum-example)
- Leg swing (pymjbase-leg-swing-example)
- Manipulator drawing (pymjbase-manipulator-drawing-example)
- Control an underactuated pendulum (pymjbase-underactuated-pendulum-example)
- Gymnast swing/release on a bar (pymjbase-gymnast-example)
- 2D Hopper (pymjbase-hopper-example)
- Initial Value Problem (pymjbase-ivp-example)
- Inverse Kinematics (pymjbase-manipulator-ik-example)
- 2D Biped (pymjbase-biped-example)

Run the command in the parenthesis to see the example, for example, to run the inverse kinematics example we can use the command:

pymjbase-manipulator-ik-example

NLOPT

I could not get the Python bindings for NLOPT to work on an M1 Mac, its only tested on Ubuntu 20.04 LTS.

Contact

If you like this repo please consider giving it a star! If you have any questions or suggestions regarding this repo you can contact me at bd1555 at nyu dot edu.