Skip to content

Latest commit

 

History

History
130 lines (94 loc) · 7.8 KB

build-from-source.md

File metadata and controls

130 lines (94 loc) · 7.8 KB
Raw

Build iDynTree from source

iDynTree is a fairly classical C++ project build with CMake, so it should be quite easy to build if you are already familiar with how you build C++ projects with CMake. If you are not familiar with the use of CMake, you can check some documentation on https://cmake.org/runningcmake/ or https://cgold.readthedocs.io .

C++ Dependencies

Before building iDynTree, you need to install the dependencies of iDynTree. Not all dependenies are strictly required, some of them can be enabled or disabled. In case they are disabled, tipically some functionality of iDynTree is not provided or disabled.

Name Strictly required? Option to enable/disable Enabled in conda-forge builds? Enabled in robotology-superbuild builds?
Eigen Yes n.a. ✔️ ✔️
Libxml2 Yes n.a. ✔️ ✔️
Assimp No IDYNTREE_USES_ASSIMP ✔️ ✔️
IPOPT No IDYNTREE_USES_IPOPT ✔️ ✔️
irrlicht No IDYNTREE_USES_IRRLICHT ✔️ ✔️
glfw No IDYNTREE_USES_IRRLICHT ✔️ ✔️
osqp-eigen No IDYNTREE_USES_OSQPEIGEN ✔️ ✔️
meshcat-cpp No IDYNTREE_USES_MESHCATCPP ❌ ❌

Install dependencies with conda-forge

If you are using conda, the dependencies of iDynTree can be installed with:

conda install -c conda-forge cmake compilers  make ninja pkg-config eigen libxml2 assimp ipopt irrlicht osqp-eigen swig python glfw

Install dependencies with apt

If you are using an apt-based distribution such as Ubuntu and you want to use apt, the dependencies can be installed via:

sudo apt-get install build-essential libeigen3-dev libxml2-dev coinor-libipopt-dev libassimp-dev libirrlicht-dev libglfw3-dev

Build

Once you installed the necessary dependencies, the iDynTree library can be compiled as any CMake based project. In the following instructions, we indicate with <additional_platform_specific_options> where you should add the platform specific options, as the use of -DCMAKE_TOOLCHAIN_FILE=[path to vcpkg]/scripts/buildsystems/vcpkg.cmake if you are using vcpkg.

With make facilities:

$ git clone https://github.com/robotology/idyntree
$ cd idyntree
$ mkdir build && cd build
$ cmake -DCMAKE_INSTALL_PREFIX=<prefix> <additional_platform_specific_options> ..
$ make
$ make install

With IDE build tool facilities, such as Visual Studio or Xcode

$ git clone https://github.com/robotology/idyntree
$ cd idyntree
$ mkdir build && cd build
$ cmake -DCMAKE_INSTALL_PREFIX=<prefix> <additional_platform_specific_options> ..
$ cmake --build . --target ALL_BUILD --config Release
$ cmake --build . --target INSTALL --config Release

If you need more help on how to build CMake-based projects, please check CGold's First step section.

In this documentation, <prefix> will indicate the installation prefix in which you installed iDynTree, i.e. the value that you passed as CMAKE_INSTALL_PREFIX during the CMake configuration.

Depending on the location in which you installed iDynTree, you may need to add <prefix>/bin to the PATH env variable and <prefix> to the CMAKE_PREFIX_PATH env variable.

Bindings

To compile bindings to iDynTree in several scriping languages, you should enable them using the IDYNTREE_USES_PYTHON, IDYNTREE_USES_LUA, IDYNTREE_USES_MATLAB, IDYNTREE_USES_OCTAVE CMake options.

Then, properly accessing bindings to iDynTree can require some additional steps.

Python

You should add to the PYTHONPATH enviromental variable the install path of the iDynTree.py file.

export PYTHONPATH=$PYTHONPATH:<prefix>/lib/python<majorPythonVersion>.<minorPythonVersion>./dist-packages/

Python (pybind11)

To compile the python bindings based on pybind11 set to TRUE the IDYNTREE_USES_PYTHON_PYBIND11 option.

NOTE: the generated bindings are not compatible with the bindings generated by SWIG. Do not expect your Python code to use either of the two without modifications.

Use

Modify your PYTHONPATH environment variable to point to the bindings installation directory:

export PYTHONPATH=${PYTHONPATH}:<prefix>/<python_package_path>

where <prefix> corresponds to the value specified in CMAKE_INSTALL_PREFIX and <python_package_path> is the Python installation prefix, as returned by

disutils.sysconfig.get_python_lib(1,0,prefix='')

for example: lib/python3.8/site-packages.

Finally to use the bindings in your Python code, simply import the package:

import idyntree.pybind as iDynTree

MATLAB

You should add to Matlab path the <prefix>/mex directory. You can modify the relative location for Matlab bindings files in the installation prefix using the IDYNTREE_INSTALL_MATLAB_LIBDIR and IDYNTREE_INSTALL_MATLAB_MFILESDIR CMake options.

Octave

You should add to Octave path the <prefix>/octave directory. You can modify the relative location for Matlab bindings files in the installation prefix using theIDYNTREE_INSTALL_OCTAVE_LIBDIR and IDYNTREE_INSTALL_OCTAVE_MFILESDIR CMake options.

MATLAB/Octave bindings modifications

All the other bindings (Python,Lua, ...) are generated by SWIG and compiled on the fly by the user, by enabling the IDYNTREE_USES_<LANGUAGE> option. The Matlab and Octave bindings are an exception because they rely on an experimental version of Swig, developed for providing Matlab bindings for the casadi project. For this reason, usually the Matlab bindigs are not generated by the users, but by iDynTree developers that have a special experimental Swig version installed. The bindings code is then committed to the repository, and the IDYNTREE_USES_MATLAB option simply enables compilation of the bindings. If you want to regenerate the Matlab bindings, for example because you modified some iDynTree classes, you can install the experimental version of Swig with Matlab support from https://github.com/robotology-dependencies/swig/ (branch matlab) and then enable Matlab bindings generation with the IDYNTREE_GENERATE_MATLAB options. For more info on how to modify the matlab bindings, see https://github.com/robotology/idyntree/blob/master/doc/dev/faqs.md#how-to-add-wrap-a-new-class-or-function-with-swig .

MATLAB/Octave high level wrappers

They are a collection of Matlab/Octave functions that wraps the functionalities of (mainly) the iDyntree class KinDynComputations into functions with a typical Matlab/Octave interface. The purpose of the high-level wrappers is to provide a simpler and easy-to-use interface for Matlab/Octave users who want to use iDyntree inside Matlab/Octave, also helping in designing code which is less error-prone and easier to debug (e.g. in case the interface of an iDyntree function will change in the future). More details and a complete list of the wrappers can be found in the wrappers README.

Usage: the wrappers package is installed together with the iDyntree bindings when compiling iDyntree with option IDYNTREE_USES_MATLAB or IDYNTREE_USES_OCTAVE set to ON. The functions can be called from Matlab/Octave using the namespace iDynTreeWrappers, i.e. iDynTreeWrappers.name_of_the_corresponding_iDynTree_method.