Using catkin_tools and wstool in a new workspace for ROS Indigo installed on Ubuntu 14.04:

source /opt/ros/indigo/setup.bash          # start using ROS Indigo
mkdir -p ~/giskard_ws/src                  # create directory for workspace
cd ~/giskard_ws                            # go to workspace directory
catkin init                                # init workspace
cd src                                     # go to source directory of workspace
wstool init                                # init rosinstall
wstool merge https://raw.githubusercontent.com/SemRoCo/giskard_examples/master/rosinstall/catkin.rosinstall
                                           # update rosinstall file
wstool update                              # pull source repositories
rosdep install --ignore-src --from-paths . # install dependencies available through apt
cd ..                                      # go to workspace directory
catkin build                               # build packages
source ~/giskard_ws/devel/setup.bash       # source new overlay

Note, the above instructions have been tested to also work under ROS Kinetic installed on Ubuntu 16.04. Just replace any occurance of indigo with kinetic.

• For a trial using on the real robot, run this command:

roslaunch /etc/ros/indigo/robot.launch                                 # on the robot
roslaunch giskard_examples pr2_interactive_markers.launch sim:=false   # on your local machine - this will start rviz too

• For a trial using iai_naive_kinematics_sim, run this command:

roslaunch giskard_examples pr2_interactive_markers.launch

Now, wait until you see the message:

Controller started.

in the console.

Note: Shall you move the rviz markers before this message is printed, TF extrapolation into the past errors will be printed in the console.

• For a trial in gazebo simulator, run these commands:

roslaunch pr2_gazebo pr2_empty_world.launch
roslaunch giskard_examples pr2_interactive_markers.launch sim:=false

Use the interactive markers to give commands to controller controlling the both arms and the torso.

Acts as a convenience interface in front of the whole_body_controller. Added values is twofold: (1) offers action interface with its feedback- and cancelation-semantics, and (2) provides intelligent parsing of motion goals to meet the specific requirements of the whole_body_controller.

• ~move (giskard_msgs/WholeBody): Movement command to be executed. The server supports only one goal at a time, and provides the following intelligent parsing of motion goals:
  • Automatic transformation of all goal poses of type geometry_msgs/PoseStamped into the reference frame of the controller using TF2.
  • Support of partial commands for body parts by using the previous commands for the respective body parts.
  • Frequent publishing of internal monitoring flags, e.g. "left arm moving" or "right arm position converged " which determine succeeded movements.

• /tf2_buffer_server (tf2_msgs/LookupTransform): Connection to TF2 buffer server to transform Cartesian goal poses.

• ~feedback (giskard_msgs/ControllerFeedback): Low-level feedback from whole_body_controller; used to determine when a movement goal succeeded.
• ~current_command_hash (std_msgs/UInt64): Hash of command currently pursued by whole_body_controller; acts as a safe-guard to synchronize whole_body_controller and controller_action_server.

• ~command (giskard_msgs/WholeBodyCommand): Command to whole_body_controller, repeated published at high frequency to kick watchdog in whole_body_controller.

Several threshold parameters determine when a motion succeeded:

• ~thresholds/motion_old (Double): Duration (in seconds) required after which a motion goal is considered old.
• ~thresholds/bodypart_moves (Double): Speed threshold (in rad/s) for the fast joint of a body part to consider that body part moving.
• ~thresholds/pos_convergence (Double) Error threshold (in m) for a position controller to have reached its goal.
• ~thresholds/rot_convergence (Double): Error threshold (in rad) for an orientation controller to have reached its goal.

A set of identifiers are used to associated the feedback from the whole_body_controller with the respective body parts.

• ~body_controllables/left_arm (List of Strings): Names of controllable variables that form the left arm of the robot.
• ~body_controllables/right_arm (List of Strings): Names of controllable variables that form the right arm of the robot.
• ~body_controllables/torso (String): Name of controllable variable making up the torso of the robot.

Finally, there are a couple of parameters which influence the behavior of the node:

• ~frame_id (String): Reference frame (known to TF) into which all Cartesian goal poses shall be transformed.
• ~update_period (Double) Time (in seconds) between updates, i.e. feedback publishes to client and commands publishes to whole_body_controller.
• ~server_timeout (Double): Time (in seconds) after which the server aborts in case it fails to initialize its action interface.