MuSe ARMCL: ARM Contact point Localization via Monte Carlo Localization

Demonstartion

A video demonstration can be found on youtube

Dependencies & Installation

To install muse_armcl the following dependencies are required. We suggest using ROS kinetic.

• ROS dependencies:

  sudo apt-get install ros-distro-orocos-kdl sudo apt-get install ros-distro-trac-ik

• External dependencies:

  • Download and install OpenMesh OpenMesh

• Internal dependencies: Create a workspace and clone the following repositories:

    git clone https://github.com/cogsys-tuebingen/muse_smc.git
    git clone https://github.com/cogsys-tuebingen/cslibs_plugins.git
    git clone https://github.com/cogsys-tuebingen/cslibs_mesh_map.git
    git clone  https://github.com/cogsys-tuebingen/cslibs_indexed_storage.git
    git clone https://github.com/cogsys-tuebingen/cslibs_kdl.git
    git clone https://github.com/cogsys-tuebingen/cslibs_utility.git
    git clone https://github.com/cogsys-tuebingen/cslibs_math.git
  

Compile with

catkin_make -DCMAKE_BUILD_TYPE=Release 

or

 catkin_make -DCMAKE_BUILD_TYPE=RelWithDebInfo	

Robot Model

You can provide your robot model as an ROS URDF model. Currently, serial manipulators with up to 3 fingers are supported. To support other kinematic trees you have to modify/update the ExternalForcesSerialChain class in cslibs_kdl. Besides, you require a mesh surface model of your manipulator in the obj-format. For surface model examples for the Kinova Jaco 2 see here.

Measurements

For measurements the particle filter requires a sensor_msgs/JointState. In this Joint State we have to provide the current joint angles and the effort field has to be set with the current external torque estimates. To estimate external torques you can use your own method or the observer provided in cslibs_kdl. if you have torque sensor in your manipulator.

Confusion Matrix Plot Script

rosrun muse_armcl plot_conf_mat.py -i <input file> -o <output file (optional)>

References

• MuSe ARMCL:

    @inproceedings{O:Zwiener:2019,
            title={{ARMCL\ - ARM Contact point Localization via Monte Carlo Localization}},
            author={Zwiener, Adrian and Hanten, Richard and Schulz, Cornelia and Zell, Andreas},
            month={October},
            year={2019},
            booktitle={2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
            note={(under review)}
    }
  

• MuSe Framework

    @INPROCEEDINGS{Hanten:2019,
            author={Hanten, Richard and Schulz, Cornelia and Zwiener, Adrian and Zell, Andreas},
            title={Multi-Sensor Integration for Sequential Monte Carlo Methods},
            booktitle={2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
            month={October},
            year={2019},
            note={(under review)},
    }