This is a reference library which uses the unitree_legged_sdk v3.2.
Python functionality is added through pybind11 and built with CMake.
This repository was created while working for the KISS Project at the Hochschule Furtwangen University.
This repository is optimized for the NX-System, that comes with the Robot.
Download the repository:
git clone --recursive https://github.com/Maddy1206/quadruped_py_control.git
Python scripts can be started already because the pre-compiled libraries are downloaded with this repository.
The Robot can be initialized in LOWLEVEL and HIGHLEVEL mode. LOWLEVEL lets you gain control over each joint independently, whereas HIGHLEVEL lets you walk the robot instantly.
If you want to run the robot with C++ programs you need to build them first. Head over to Building.
- Boost (version 1.5.4 or higher)
- CMake (version 2.8.3 or higher)
- LCM (version 1.4.0)
- g++ (version 7.5.0 or higher)
- pybind11 (version 2.10.4 or higher)
- Python (version 3.6)
Install Boost, CMake, the latest 8.x version of g++ and the latest version of Python 3.6 with the following command:
sudo apt-get update
sudo apt-get install libboost-all-dev make cmake g++-7 gcc-7 python3.6
Download LCM and unzip to /home/[username] and build the software:
cd lcm-1.4.0
mkdir build
cmake ..
make
sudo make install
You must download the pybind11 library seperately and put it in the specific folder.
Make sure the path to the pybind11 files will be /quadruped_py_control/python_wrapper/third-party/pybind11/. Copy all files from the download to the pybind11-folder in this repository or use git (if you used the --recursive flag, the pybind11 repo is already in the folder).
cd ~/quadruped_py_control/python_wrapper/third-party
git clone https://github.com/pybind/pybind11.git
The source code of the pyrealsense2 library has not been added to this repo because the compiling process takes longer than 30 minutes. Instead, the pre-compiled libraries have been added to /lib/python/.
If you cannot find git, run:
sudo apt-get install git
If you cannot find msgpack.hpp while compiling, run:
sudo apt-get install libmsgpack*
In python_interface.cpp the exposure of C++ classes and types is declared with the help of pybind11.
In both CMakeLists.txt the compiler standard has been set to 14 due to compatibility reasons with pybind11.
In both CMakeLists.txt make sure that the correct architecture type is selected! amd for RaspberryPi and VirtualMachine and arm when working with the NVIDIA NX!
libunitree_legged_sdk_amd64.so or libunitree_legged_sdk_arm64.so
Download the quadruped_py_control library (and unzip) to /home/[username] and build with
cd quadruped_py_control
mkdir build
cd build
cmake ..
make
The C++ files should be written to the build-folder and the CPython library to the ../lib/python folder.
Run the C++ commands with
sudo ./example_walk
and the Python commands with
sudo python3 example_walk.py
in the respective folder.
When working with an IDE (preferably Visual Studio Code) make sure to set the C++ compiler and the Python interpreter to the respective version!