Implementation of an approximate inverse kinematics for robots described by URDF files in C++ with Python bindings. The approximate IK allows you to set a weight for the position and the orientation to define how important it is to reach them. Typically it is more important to reach the position exactly than the orientation. Sometimes there are situations where it is not even possible to reach the position at all while ignoring the orientation. In these cases we want to find the IK solution that is closest to the desired goal.

We achieve these features by defining a cost function that has a weight for the position and the orientation constraint respectively and lets the choice of the weight for the user. We solve the IK problem by optimizing the cost function with respect to the joint angles with L-BFGS-B, a local optimizer that respects boundaries (i.e. joint limits). The required derivatives are calculated numerically which is usually fast enough.

Install the dependencies and install the C++ header:

mkdir build
cd build
cmake ..
make install

Now you can use

#include <approxik/approxik.hpp>

in your project.

./get_dependencies.sh
mkdir build
cd build
cmake ..
cd ../python
sudo python setup.py install

Note that dependencies will be installed in the subfolders deps/install. Source the env.sh to set the correct environment variables.

• approxik: C++ implementation (in one header file)
• cmake: additional CMake modules
• data: contains data files (e.g. URDFs, trajectories)
• deps: dependencies
• evaluation: scripts to generate plots for the paper
• examples: several scripts for visualization of the approximate IK approach
• paper: summary of the method, experiments, etc.
• prototype: first prototype of the approximate IK
• python: Python wrapper for C++ implementation
• src: C++ example program
• test: unit tests