About

This repository contains the Zig bindings and implementations for the MuJoCo C API. Plus a lot of additions to make it a full blown Robot Training Framework.

Zig version 0.15.x

MuJoCo version 3.3.8

https://github.com/user-attachments/assets/df516e11-f744-4ea8-813b-1cb99de78644

Robotic Dog

https://github.com/user-attachments/assets/398d0165-595a-4706-9103-7d2417a66357

2D Self-Learning Agent for playing the snake game being used in a 3D enviroment - using DQNAgentC C framework I am working on. All the code will be here soon!

Working on a self-driving agent that can be loaded by a human & unloads by itself - something like a parking / unloading mode when the target is reached.

Why phyzx?

I saw the Jolt physics engine bindings which are called zphysics and I thought what a wasted opportunity. This is trying to be more like the official python bindings & the dm_control python package rather than just a physics engine / MuJoCo wrapper, BTW.

Due to both C and Zig being very low level / lightweight languages it should make this more accessible to run simulations on CPU-only & lower spec hardware in general! Also more importantely once you actually want to convert the simulation's inputs and outputs into some actual system you can use more or less the same code just redefine things a bit.

Getting Started

You can call on the phyzx-xyz building block modules directly or redefine them or even the src/phyzx-api.zig entirely to expose the functions for your needs & workflow!

You should be able to just:

git clone https://github.com/lab-key/phyzx.git
# Everything in the examples directory gets built automatically
zig build run-control_arm26 # Or ./zig-out/bin/load_arm26 - Just that until I integrate everything properly...

This repository uses my zmujoco repository as a git submodule all the files are already fetched & everything is already setup. So you don't need the setup.zig & cleanup.zig In case you want to setup your own zmujoco / mujoco repositories. Check the `examples/local' directory use it as a general template and adjust it for your needs, filepaths etc.

Current and Future plans / issues - TODO

A lot of potential due to Zig being interoberable with C so I'm not sure if I should maybe even keep it ' mixed and dirty ' a bit and make examples on how to do it yourself or try to keep it pure Zig although that would be kinda redundant as Zig was designed with this feature, as C is an old language so..

First of all I'm probably not aware of everything that exists and second reprogramming and wrapping everything properly is just too time consuming. Also some of the A.I. / control stuff I'm testing is using OpenBLAS which you can't build directly with Zig as it uses Fortran for example.. So it's either making a Fortran compiler in Zig - using it as a module ( might not even be possible ), invoking the make command which is an external dependency ( maybe making that into a Zig module ) or just having a system-wide install of OpenBLAS as a dependency which is what I do internally but I don't have time to test everything everywhere properly there might be versioning issues especially across different OSes and I don't want to have the experimental stuff break something for someone - 'works on my machine' type of thing. So I decided to add an examples/experiment directory and I'll make some docs on how to use it as well. I will place the build.zig files separately and then you combine / build based on what you need.

I'll see depends on the interest as well.

A proper GUI Editor and things mapped out for it but want more ML integration as well otherwise it won't have the same charm....
Examples, Tests setting everything for v0.0.1 properly
A nice clean way to integrate or point to C libraries that are not buildable with Zig at all. 2.1 Adding 3rd party modelfiles
All the demos and most things I am internally working on are done very ' dirty ' with direct C interop need to either properly wrap it or reprogram it - possibly start working on the experimental directory..

Documentation

All the documenatation is in the docs/ directory.

I highly recommend you look at the tests directory as well as it contains a lot of examples as does the zmujoco project.

MuJoCo's documentation can be found at mujoco.readthedocs.io. Upcoming features due for the next release can be found in the changelog in the "latest" branch.

Citation

If you use MuJoCo for published research, please cite:

@inproceedings{todorov2012mujoco,
  title={MuJoCo: A physics engine for model-based control},
  author={Todorov, Emanuel and Erez, Tom and Tassa, Yuval},
  booktitle={2012 IEEE/RSJ International Conference on Intelligent Robots and Systems},
  pages={5026--5033},
  year={2012},
  organization={IEEE},
  doi={10.1109/IROS.2012.6386109}
}

License and Disclaimer

ReStructuredText documents, images, and videos in the doc directory are made available under the terms of the Creative Commons Attribution 4.0 (CC BY 4.0) license. You may obtain a copy of the License at https://creativecommons.org/licenses/by/4.0/legalcode.

Source code is licensed under the Apache License, Version 2.0. You may obtain a copy of the License at https://www.apache.org/licenses/LICENSE-2.0.

This is not an officially supported Google product.

Dependencies

README