PIPER

PIPER (Probabilistic Inference based Platform for Essential problems in Robotics) is a modular package that provides support for algorithms that use probabilistic inference on factor graphs to solve various robotics problems. Each module can be independently installed and implements an easy ROS interface for that algorithm to run on any simulated or real robots. Currently PIPER supports the following algorithms:

• GPMP2 - Gaussian Process Motion Planner 2
• STEAP - Simultaneous Trajectory Estimation and Planning
• more coming soon ...

PIPER is being developed by Mustafa Mukadam at the Georgia Tech Robot Learning Lab. See documentation for information on usage and development.

Table of Contents

• Prerequisites
• Compilation and Installation
• Questions and Bug reporting
• Citing
• FAQs
• License

Prerequisites

• Install ROS. We have tested up to ROS Kinetic on Ubuntu 16.04.
• Install GPMP2 C++ library.

Compilation and Installation

• Initialize a catkin workspace (if you are using an existing catkin workspace this step is not needed)

  mkdir -p ~/piper_ws/src
  cd ~/piper_ws/src
  catkin_init_workspace
  

  Before running setup the environment variables

  source ~/piper_ws/devel/setup.bash
  

• Clone this repository in ~/piper_ws/src

  git clone https://github.com/gtrll/piper.git
  

• To compile only the piperbase library, in the catkin workspace directory do

  catkin_make
  

• Otherwise, to install some module, for example, X: first make sure to install dependencies for X, besides the prerequisites for piperbase, then do

  catkin_make -build_flag_X
  

• Similarly, to install multiple modules, for example, X and Y: install all their dependencies and then use their appropriate flags together

  catkin_make -build_flag_X -build_flag_Y
  

  See table below for currently supported modules, their dependencies and build flags

  Module	Other Dependencies	Build Flag
  GPMP2	None	-DBUILD_GPMP2_INTERFACE:OPTION=ON
  STEAP	None	-DBUILD_STEAP_INTERFACE:OPTION=ON
  ALL	All from above	-DBUILD_ALL_INTERFACE:OPTION=ON

Questions and Bug reporting

Please use Github issue tracker to report bugs. For other questions please contact Mustafa Mukadam.

Citing

If you use PIPER in an academic context, please cite any module/algorithm specific publications you use, and cite the following:

@article{mukadam2017piper,
  title={{PIPER}},
  author={Mukadam, Mustafa},
  journal={[Online] Available at \url{https://github.com/gtrll/piper}},
  year={2017}
}

FAQs

• Q: Cannot find trajectory_control server error

  A: To be able to run the included examples you need to have an action server set up in your real or simulated robot's API. See trajectory_control_topic in robot config.

• Q: Node crashes without running anything

  A: First verify if the error is related to different versions of Eigen being linked. Run gdb ~/piper_ws/devel/lib/piper/gpmp2_interface and r. If you get the following error Program received signal SIGSEGV, Segmentation fault. 0x00007ffff672a2e7 in gtsam::noiseModel::Constrained::Constrained(Eigen::Matrix<double, -1, 1, 0, -1, 1> const& Eigen::Matrix<double, -1, 1, 0, -1, 1> const&) () from /usr/local/lib/libgtsam.so.4, it means that there are at least two versions of eigen conflicting. Make sure to install GTSAM with system Eigen. Then install GPMP2 and try this again.

License

PIPER is released under the BSD license. See LICENSE file in this directory.