gaolongsen/mujoco_mpc

Real-time behaviour synthesis with MuJoCo, using Predictive Control

MuJoCo MPC (MJPC) is an interactive application and software framework for real-time predictive control with MuJoCo, developed by DeepMind.

MJPC allows the user to easily author and solve complex robotics tasks, and currently supports three shooting-based planners: derivative-based iLQG and Gradient Descent, and a simple yet very competitive derivative-free method called Predictive Sampling.

• Overview
• Graphical User Interface
• Installation
  • macOS
  • Ubuntu
  • Build Issues
• Predictive Control
• Contributing
• Known Issues
• Citation
• Acknowledgments
• License and Disclaimer

Overview

To read the paper describing this software package, please see out preprint.

For a quick video overview of MJPC, click below.

For a longer talk at the MIT Robotics Seminar describing our results, click below.

Graphical User Interface

For a detailed dive of the graphical user interface, see the MJPC GUI documentation.

Installation

You will need CMake and a working C++20 compiler to build MJPC. We recommend using VSCode and 2 of its extensions (CMake Tools and C/C++) to simplify the build process.

1. Clone the repository: git clone https://github.com/deepmind/mujoco_mpc.git
2. Configure the project with CMake (a pop-up should appear in VSCode)
3. Build and run the mjpc target in "release" mode (VSCode defaults to "debug"). This will open and run the graphical user interface.

macOS

Additionally, install Xcode.

Ubuntu

1. if you don't have any issues for download problem, follow the installation, step 1

2. Go into the main project, open the terminal and then use camke . to compile the project.

3. You need to wait for a while and ignore any warnings or errors from the terminal.

4. After the last step finished, you need to change several files to avoid some fatal and suck errors when you compile the whole probject:

   1. Go to mujoco_mpc/cmake/MpcOptions.cmake and /home/wam/mujoco_mpc/_deps/mujoco-build/CMakeFiles/mujoco.dir/flags.make, Comment or remove all -Werror variable to make sure even there are some warnings happen, you can also compile the project.
   2. Go to mujoco_mpc/mjpc/agent.cc and add (struct mjuiDef_) in line 366, line 386 and line 389 to make sure that no errors occur during the assignment of structure variables.

   (Note: If you use officiial version, strongly suggest follow step 4; If you use my version, git my branch directly because I have done everything for that.)

5. Then you input cmake --build . on the terminal and need to wait for a long while until the project compile finishing. (around 10 min in my PC).

Additionally, install:

sudo apt-get install libgl1-mesa-dev libxinerama-dev libxcursor-dev libxrandr-dev libxi-dev ninja-build

Build Issues

If you encounter build issues, please see the Github Actions configuration. Note, we are using clang-14.

Predictive Control

See the Predictive Control documentation for more information.

Contributing

See the Contributing documentation for more information.

Known Issues

MJPC is not production-quality software, it is a research prototype. There are likely to be missing features and outright bugs. If you find any, please report them in the issue tracker. Below we list some known issues, including items that we are actively working on.

• We have not tested MJPC on Windows, but there should be no issues in principle.
• Task specification, in particular the setting of norms and their parameters in XML, is a bit clunky. We are still iterating on the design.
• The Gradient Descent search step is proportional to the scale of the cost function and requires per-task tuning in order to work well. This is not a bug but a property of vanilla gradient descent. It might be possible to ameliorate this with some sort of gradient normalisation, but we have not investigated this thoroughly.

Citation

If you use MJPC in your work, please cite our accompanying preprint:

@article{howell2022,
  title={{Predictive Sampling: Real-time Behaviour Synthesis with MuJoCo}},
  author={Howell, Taylor and Gileadi, Nimrod and Tunyasuvunakool, Saran and Zakka, Kevin and Erez, Tom and Tassa, Yuval},
  archivePrefix={arXiv},
  eprint={2212.00541},
  primaryClass={cs.RO},
  url={https://arxiv.org/abs/2212.00541},
  doi={10.48550/arXiv.2212.00541},
  year={2022},
  month={dec}
}

Acknowledgments

The main effort required to make this repository publicly available was undertaken by Taylor Howell and the DeepMind Robotics Simulation team.

License and Disclaimer

All other content is Copyright 2022 DeepMind Technologies Limited and licensed under the Apache License, Version 2.0. A copy of this license is provided in the top-level LICENSE file in this repository. You can also obtain it from https://www.apache.org/licenses/LICENSE-2.0.

This is not an officially supported Google product.