gpmp2

Note: Version compatible with latest GTSAM is being maintained at borglab/gpmp2.

GPMP2

This library is an implementation of GPMP2 (Gaussian Process Motion Planner 2) algorithm described in Motion Planning as Probabilistic Inference using Gaussian Processes and Factor Graphs (RSS 2016). The core library is developed in C++ language with an optional Python 2.7 toolbox. GPMP2 was started at the Georgia Tech Robot Learning Lab, see THANKS for contributors.

Prerequisites

• CMake >= 3.0 (Ubuntu: sudo apt-get install cmake), compilation configuration tool.
• Boost >= 1.50 (Ubuntu: sudo apt-get install libboost-all-dev), portable C++ source libraries.
• Anaconda2, virtual environment needed if installing python toolbox.
• GTSAM == wrap_export, a C++ library that implements smoothing and mapping (SAM) framework in robotics and vision. Here we use the factor graph implementations and inference/optimization tools provided by GTSAM.

Installation (C++ only)

• Install GTSAM.

  git clone https://github.com/borglab/gtsam.git
  cd gtsam
  git checkout wrap-export
  mkdir build && cd build
  cmake ..
  make check  # optional, run unit tests
  sudo make install
  

• Setup paths.

  echo 'export LD_LIBRARY_PATH=/usr/local/lib:${LD_LIBRARY_PATH}' >> ~/.bashrc
  echo 'export LD_LIBRARY_PATH=/usr/local/share:${LD_LIBRARY_PATH}' >> ~/.bashrc
  source ~/.bashrc
  

• Install gpmp2.

  git clone https://github.com/gtrll/gpmp2.git
  cd gpmp2 && mkdir build && cd build
  cmake ..
  make check  # optional, run unit tests
  sudo make install
  

Installation (C++ with Python toolbox)

• Setup virtual environment.

  conda create -n gpmp2 pip python=2.7
  conda activate gpmp2
  pip install cython numpy scipy matplotlib
  conda deactivate
  

• Install GTSAM.

  conda activate gpmp2
  git clone https://github.com/borglab/gtsam.git
  cd gtsam
  git checkout wrap-export
  mkdir build && cd build
  cmake -DGTSAM_INSTALL_CYTHON_TOOLBOX:=ON ..
  make check  # optional, run unit tests
  sudo make install
  conda deactivate
  

• Setup paths.

  echo 'export LD_LIBRARY_PATH=/usr/local/lib:${LD_LIBRARY_PATH}' >> ~/.bashrc
  echo 'export LD_LIBRARY_PATH=/usr/local/share:${LD_LIBRARY_PATH}' >> ~/.bashrc
  echo 'export PYTHONPATH=/usr/local/cython:${PYTHONPATH}' >> ~/.bashrc
  source ~/.bashrc
  

• Install gpmp2.

  conda activate gpmp2
  git clone https://github.com/gtrll/gpmp2.git
  cd gpmp2 && mkdir build && cd build
  cmake -DGPMP2_BUILD_PYTHON_TOOLBOX:=ON ..
  make check  # optional, run unit tests
  sudo make install
  cd ../gpmp2_python && pip install -e .
  conda deactivate
  

Citing

If you use GPMP2 in an academic context, please cite following publications:

…

License

GPMP2 is released under the BSD license, reproduced in LICENSE.