kincpp

kincpp is a C++ library for computing both forward and inverse kinematics for arbitrary robot manipulators. It is written entirely in C++ for speed and has pybind bindings for python. Any manipulator can be modelled by simply providing its home configuration end-effector transform M and the joint screw axes S. A tool to compute this for modelled robots can be found here.

Core logic to compute FK and Newton-based IK was largely extracted / refactored from ModernRoboticsCpp (a C++ implementation of ModernRobotics). Constraint-based IK using a quadratic programming approach is also available and was ported over to C++ from robotics-toolbox-python. Tentative profiling results show an order of magnitude speed increase over the original Python version.

Build from Source

To build from source, first install the necessary dependencies Eigen and pybind.

sudo apt install libeigen3-dev pybind11-dev

Then, simply build the shared library file

mkdir build && cd build && cmake ..
make -j4

This will produce a kincpp.cpython-{YOUR_PYTHON_VERSION}-{YOUR_ARCH}.so file that you can use directly for your python applications. You can also install it as python package by running

pip install .

A static library is also created which can be linked to in cmake using the target kincpp_lib.

Testing

Once kincpp is installed in some way, you can test that everything is working by running the provided unit test.

python testing/test_kinematics.py
python testing/ik_qp_test.py

These files also contain useful examples for instantiating a KinematicsSolver and calling the forward and inverse kinematics functions.

Troubleshooting

If your system cannot find Eigen, you may need to create symlinks if your Eigen package is actually named eigen3/Eigen. Simply locate your Eigen installation location, locate eigen, and then create the necessary symlinks. For example, if Eigen is found in /usr/include, then you would do the following:

cd /usr/include
sudo ln -sf eigen3/Eigen Eigen