This package contains configuration files and scripts for bringing up the LCSR Barrett WAM robot interfaces both on real hardware and in simulation with controllers for various tasks.

This package uses the following platforms:

• ROS The Robot Operating System
• Orocos The Open Robot Control Software
• rtt_ros_integration Tools for using Orocos Real Time Toolkit (RTT) and ROS together

For real and simulated hardware interfaces, it uses the following packages:

• barrett_model An URDF model of the Barrett WAM and BHand
• orocos_barrett Orocos interfaces to the real and simulated Barrett WAM robot and BHand

The controllers, built as Orocos RTT components, can be found in the following packages:

• conman A dynamically-switching controller manager architecture
• lcsr_controllers A collection of robot-agnostic controllers which use the above frameworks

When running the arm in simulation, we also use the following tools:

• Gazebo A modular physical 3D simulation framework
• rtt_gazebo Tools for running Orocos RTT components in Gazebo
• gazebo_ros_pkgs Tolls for integrating Gazebo and ROS

System Packages (not handled by rosdep):

sudo apt-get install git python-pip curl google-mock ros-hydro-cmake-modules
sudo apt-get install omniorb libomniorb4-dev omniidl omniorb-nameserver

Catkin Tools:

sudo pip install catkin_tools

Packages required to build gazebo:

sudo apt-get install libtar-dev protobuf-compiler libprotobuf-dev

Barrett-Patched LibConfig++:

echo "export OROCOS_TARGET=xenomai" >> ~/.bashrc
export XENOMAI_ROOT_DIR=/usr/xenomai
wget http://web.barrett.com/support/WAM_Installer/libconfig-barrett_1.4.5-1_`dpkg --print-architecture`.deb
sudo dpkg -i libconfig-barrett_1.4.5-1_`dpkg --print-architecture`.deb

First, decide if you want to install the simulation workspace or the hardware workspace:

export SIM_OR_HW=sim

or:

export SIM_OR_HW=hw

Then, create, clone, and build the workspace:

# make the workspace
mkdir -p barrett_ws/src
cd barrett_ws
# initialize catkin
catkin config --init --cmake-args -DENABLE_CORBA=ON -DCORBA_IMPLEMENTATION=OMNIORB
# initialize wstool source checkouts
cd src
wstool init
curl https://raw.githubusercontent.com/jhu-lcsr/rosinstalls/master/barrett/wam_common.rosinstall | wstool merge -
curl https://raw.githubusercontent.com/jhu-lcsr/rosinstalls/master/barrett/wam_$SIM_OR_HW.rosinstall | wstool merge -
wstool update -j4
# install the system dependencies for source packages
rosdep install --from-paths . --ignore-src
# build the workspace
catkin build

The entry point for any experiments and examples in this repository is a ROS launchfile. Any application should be launchable either on the real robot or in simulation. These launchfiles load the following:

• An Orocos Deployer which connects to real hardware or hardware in simulation;
• An Orocos Script (.ops file) which loads and configures a graph of Orocos RTT components to control the arm and hand;
• The ROS parameters for any controllers that were loaded by the specified Orocos script;
• A robot_state_publisher to publish the TF frames for the Barrett devices to ROS;
• When simulating the arm, it will also start an instance of the Gazebo Simulator via the gazebo_ros_pkgs tools for running Gazebo with ROS.

When working with the real hardare, the aforementioned Orocos Script is specified in the launchfile and is loaded directly into an Orocos deployer, but when working in simulation, the Orocos Script is loaded by the rtt_gazebo_deployer plugin in the Gazebo gzserver process.

These orocos scripts do the following:

• import RTT plugins from other ROS packages
• create RTT components defined in the imported RTT plugins
• connect RTT data ports of the various components
• connect RTT data ports to ROS topic streams
• enable and disable the initial set of RTT components

Once the system is launched, a user interacts with it entirely via ROS interfaces.

This package requires all repositories listed in the lcsr_wam or lcsr_wam_sim LCSR Rosinstall depending on whether you intend to use the real or simulated hardware:

• Real hardware: https://github.com/jhu-lcsr/rosinstalls/blob/master/applications/lcsr_wam.rosinstall
• Simulateed hardware: https://github.com/jhu-lcsr/rosinstalls/blob/master/applications/lcsr_wam_sim.rosinstall

You can command the manual trapezoidal trajectory generator from ROS like so:

Home position:

rostopic pub -1 /barrett/man_traj/joint_position_cmd trajectory_msgs/JointTrajectoryPoint "{ positions: [0., -1.57, 0.00, 3.0, 0.0, -0.8, 0.0] }"

rostopic pub -1 /barrett/man_traj/joint_position_cmd trajectory_msgs/JointTrajectoryPoint "{ positions: [-0.4, -1.5, 0.00, 1.5, -1.2, 1.0, 1.8] }

You can command the gripper over ros as well:

rostopic pub  /hand_cmd oro_barrett_msgs/BHandCmd "{ cmd: [2.0, 2.0, 2.0, 0.0], mode: [2,2,2,1] }" 

The following are notes on simulation realtime performance for various machines and configurations:

• Intel Core i7-3520M
  • ODE
    • Single WAM arm: 99%
    • Single WAM arm + BHand: 92% (open) 82% (grasping)

• lcsr_controllers 0.4.0
• barrett_model 0.4.0