DualManipulation

This repository is a container for all packages needed for the dualManipulation task. To download every required package at once use the following command:

git clone --recursive https://github.com/CentroEPiaggio/dual_manipulation.git

Refer to this page for setting up an ssh-key connected to your GitHub account.

Dependencies

This package requires the repository vito-robot (branch: soma_devel) and its dependencies (kuka-lwr and pisa-iit-soft-hand): follow the instructions at the vito-robot page.

For a real scenario with the use of vision, you might need to perform camera-robot calibration. This calibration package can help you to do that..

Updating submodules

When you clone this repo you will have all the submodules fixed at a certain version (that can be updated just committing on this repo when you update the submodule). First of all checkout the desired branch of dual_manipulation:

cd dual_manipulation
git checkout <branch-name>
git submodule update --remote

This will only update the branch registered in the .gitmodule, and by default, you will end up with a detached HEAD. Now you can catkin_make the packages to try out the planner.

If you wish to contribute to the project, befor editing the code of a <branch> in <package-folder>, you should run the following:

cd dual_manipulation/<package-folder>
git checkout <branch>
git pull

Documentation

To generate the documentation of the project you have to run the following series of commands:

cd dual_manipulation
doxygen documentation.doxygen

and then open the file index.html in the generated html folder with a browser.

Running a DualManipulation Demo

A full demo can be run with only kinematics (moveit and rviz only), simulation, or real robot. The main difference is the launch command needed to load either of those possibilities.

Let's explore useful parameters in more details

Configurations

Change in dualmanipulation/ik_control/config/${ROBOT_NAME}_ik_control.yaml file the parameter

kinematics_only: true

in order to avoid waiting for controller responses, i.e. when using kinematics only execution, and

kinematics_only: false

when using simulation or the real robot.

Change in dualmanipulation/shared/config/${ROBOT_NAME}_dual_manipulation.yaml

use_vision: false

to allow for positioning the source object via the GUI.

Kinematics only launch

The planning procedure can be tested and, assuming controllers will do their job, the whole demo can be performed.

Load the robot model, moveit and rviz using the command

roslaunch vito_moveit_configuration demo.launch

Simulation launch

Alternatively, the planning procedure can be tested also considering response from the controllers, assuming controllers will work similarly in the simulated and real robot; the whole demo can be performed, apart from not having the object in the simulated scene.

Load the simulated robot, moveit and rviz using the command

roslaunch vito_description display.launch

Real robot launch

In this way, the physical robot can be used.

• roslaunch vito_description display.launch use_robot_sim:=false load_moveit:=false right_arm_enabled:=true left_arm_enabled:=true right_hand_enabled:=true left_hand_enabled:=true use_rviz:=false
• roslaunch vito_moveit_configuration pacman_demo.launch

Common procedure

Once the robot_model has been loaded in one or the other way, and parameters have been set appropriately, THREE more shells are needed for dualmanipulation. The commands to run are

• rosrun dual_manipulation_ik_control dual_manipulation_ik_control
• rosrun dual_manipulation_state_manager dual_manipulation_state_manager
• rosrun dual_manipulation_gui dual_manipulation_gui

Then, after everything is steady (ik_control might take some time to load) you can start the experiment using the GUI, where the sequence of actions to be performed is

• select an object (if using the GUI for setting the target), publish markers and move them where source and target should be in the experiment (only target if the vision is used)
• get info button: information will be given about which objects have been found in the scene
• set target button: to be pressed after selecting the target position for the object; a graph will be shown in the GUI representing all possible states and transitions of the high level planner
• plan button: a tentative plan will be computed, and state_manager will check for kinematics feasibility; few iterations could be performed before finding a viable option from source to target; a red line will show the planned high level sequence of actions
• after a plan has been found, use start moving button to iterate between cartesian planning and command execution, which will all be automated.

If, for any reason, you want the robot to stop at a certain point, press the stop button and the trajectory will halt; you can also use the home function to move the whole robot back home.

exit closes the state machine, reset clears all states and allow for a fresh start. abort plan can be used if a high-level planned path is not what we wanted, to plan again (if no condition is changed, anyway, the planning result will be the same!), and abort move is used to abort planning-moving for the robot, without interrupting the current motion.