Contents

This contains two examples of OCP solution with casadi Opti class. As a general documentation reference for casadi, you can refer to the official doc and look at the many examples

In particular:

• cartpole.py uses a custom made problem with analytic dynamics, so it's a standalone example
• doublependulm.py uses pinocchio.casadi module (requires pinocchio >= 3.0) to discretize the integrator dynamics with the algorithms implemented in pinocchio

Similarly as shown in the code the functions in pinocchio.casadi can be used to impose costs or constraints.

Set up

This package relies on Pinocchio-3x (internal repo @Inria), which in turns rely on several packages.

APT packages

Some packages rely on robotpkg APT repo, that can be set up following Pinocchio installation procedure. If following this procedure, don't forget to also set up your environment variables.

You need the following packages:

sudo apt install cmake  cmake-curses-gui  ccache  doxygen  libgmp10-dev  libmpfr-dev  robotpkg-py38-casadi libsdformat-dev graphviz libboost-timer-dev robotpkg-py38-pinocchio

HPP-FCL devel branch

You need to recompile HPP-FCL. Clone the repo, then execute:

cd hpp-fcl
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=RELEASE -DCMAKE_INSTALL_PREFIX=$HOME/compil/pinocchio
make -sj8 install

Add the following to your environment:

export PINOCCHIO_INSTALL_DIR=$HOME/compil/pinocchio
export PATH=${PINOCCHIO_INSTALL_DIR}/bin:${PATH}
export PKG_CONFIG_PATH=${PINOCCHIO_INSTALL_DIR}/lib/pkgconfig:${PKG_CONFIG_PATH}
export PKG_CONFIG_PATH=${PINOCCHIO_INSTALL_DIR}/lib64/pkgconfig:${PKG_CONFIG_PATH}
export LD_LIBRARY_PATH=${PINOCCHIO_INSTALL_DIR}/lib:${LD_LIBRARY_PATH}
export LD_LIBRARY_PATH=${PINOCCHIO_INSTALL_DIR}/lib64:${LD_LIBRARY_PATH}
export PYTHON_LOCAL_PATH=lib/python3/dist-packages
export PYTHONPATH=${PINOCCHIO_INSTALL_DIR}/${PYTHON_LOCAL_PATH}:${PYTHONPATH}

Check by loading the hpp-fcl module in python, it should point to your compilation folder:

python -c "import hppfcl; print(hppfcl.__file__)"

Compile Pinocchio-3x

Clone the Inria repo of Pinocchio

git clone --recursive git@gitlab.inria.fr:jucarpen/pinocchio.git -b topic/template-instantiation

Compile it with the following options:

cd pinocchio
mkdir build
cd build
cmake .. -DBUILD_PYTHON_BINDINGS_WITH_BOOST_MPFR_SUPPORT=OFF -DBUILD_WITH_CASADI_SUPPORT=ON -DBUILD_WITH_COLLISION_SUPPORT=ON -DCMAKE_BUILD_TYPE=RELEASE -DCMAKE_INSTALL_PREFIX=$PINOCCHIO_INSTALL_DIR
make -sj8 install

Viewer

You might consider to install Gepetto viewer

sudo apt install robotpkg-py38-qt5-hpp-gui

Don't forget to start the viewer before executing the scripts needing it (otherwise you will typically get the following error "AttributeError: 'GepettoVisualizer' object has no attribute 'viewer'"):

gepetto-viewer

Alternatively, you will have to comment out the display command in the example scripts.

Git commits

The installation procedure as been check with HPP-FCL commit 39fde9cb and Pinocchio commit 74deb9ab.

Check

You can run python -c "import pinocchio.casadi as cpin; print(cpin.__file__)"