xcub-moveit2

Maintainer

	@martinaxgloria

Overview

This repository contains the current state of the ROS 2 packages for using iCub and ergoCub with MoveIt.

Assumptions

These packages were generated and tested with ros humble distro on a Ubuntu 22.04 machine.

Prerequisites and dependencies

First of all, install ROS 2 Humble on your machine and configure your ROS 2 environment. Then, follow the MoveIt 2 install guide to build all the necessary dependencies.

Moreover, it is mandatory to install yarp-devices-ros2 on your machine to use custom ROS messages and services defined in yarp_control_msgs. To do this, you can follow the installation procedure described in the repository. Otherwise, starting from the distro v2024.11.0, it can be compiled within the robotology-superbuild by enabling the ROBOTOLOGY_USES_ROS2 CMake option .

This repository contains some devices and custom ROS 2 interfaces with different purposes and, in particular for this application, it contains the possibility to control a yarp-based robot with ROS 2. To enable this features, you have to add the msgs_name parameter in your configuration file that inizializes the device controlBoard_nws_ros2, for example:

Finally, TRAC-IK is chosen as inverse kinematics solver. It is more accurate and faster when dealing with complex kinematic chains with respect to KDL Kinematics, which represents the standard for MoveIt 2. To install it inside your ROS 2 workspace:

cd ~/<ros2_ws>/src
source /opt/ros/humble/setup.bash

# Clone the repository inside your ros2 workspace and build it
git clone https://bitbucket.org/traclabs/trac_ik.git -b rolling-devel
cd trac_ik
colcon build
source install/setup.bash

NOTE: If you don't want to source your ros humble setup file each time a new shell is open, you can add this command in your .bashrc:

echo "source /opt/ros/humble/setup.bash" >> ~/.bashrc

Install

To install the project from source, please follow only one of the following two sections.

Install with colcon

In your ROS 2 workspace, clone the repo:

cd ~/<ros2_ws>/src
git clone https://github.com/icub-tech-iit/xcub-moveit2

and then build the environment:

cd ..
source /opt/ros/humble/setup.bash 
colcon build
source install/setup.bash

Install as a single CMake project

If you want to build this repository as a single CMake project, you can use the CMakeLists.txt provided in xcub_moveit_all_packages:

git clone https://github.com/icub-tech-iit/xcub-moveit2/
cd xcub-moveit2/xcub_moveit_all_packages
cmake -Bbuild -S. -DCMAKE_INSTALL_PREFIX=<install_prefix>
cmake --build build
cmake --install build

# Make ROS configuration files available in [ament index](https://github.com/ament/ament_index)
export AMENT_PREFIX_PATH=$AMENT_PREFIX_PATH:<install_prefix>

Usage on real hardware

If you want to control a yarp-based robot (like iCub) using the proposed framework, you have to make sure that your machine is in communication with the robot one on which ros2 is installed. For this reason, be sure that your laptop and the robot machine are on the same network:

• configure your yarpserver as to be the one running on the robot environment;
• after checking your ip address and the robot one, customize the cyclonedds.xml file in this repository as follow:

• export this configuration file in the .bashrc in order to be used by Cyclone DDS:

export CYCLONEDDS_URI=/<path/to/your/ros2_ws>/src/xcub-moveit2/cyclonedds.xml

• repeat the last two steps on the robot machine, changing the ip address accordingly.

However, if you want to work in simulation instead of on the real hardware, you can skip the list above.

Package description

This section aims to give a brief description of what each package contains.

xcub_ros2_controllers

This package contains xcub_ros2_controllers plugin that is used in ros2_control framework. It includes:

• a position state interface used to read the position of each joint;
• a velocity state interface used to read the velocity of each joint;
• a position command interface used to forward the desired position to the joints.

xcub_moveit_robot

This package contains some launch files, depending on the nodes you want to run.

• robot.launch.py: it allows to start up a set of nodes to bringup the robot. It spawns the model on rviz, publishes the state of the robot to tf2 topic and starts the move_group node, which provides some ROS actions and services for using MoveIt 2 with your robot.
• robot_sim.launch.py: it's the same as before, but it spawns the model also in gazebo environment to work with the simulated robot.
• robot_controls.launch.py: this launch file allows to run the controller manager node for ros2_control and the nodes for the single controllers (one for each part).
• circle_demo.launch.py and grasp_demo.launch.py: as the name suggests, they are two examples of commanding the robot in the cartesian space using torso + right_arm as planning group.

Before running the nodes, make sure that your environment variable YARP_ROBOT_NAME is properly set with the name of your robot. If not, set it for each shell according to the chosen model, for example:

export YARP_ROBOT_NAME="iCubGenova11"

or, for simulated models:

export YARP_ROBOT_NAME="iCubGazeboV2_5"

icub_moveit_config

This package contains the configuration files to make iCub working with MoveIt2. In particular, each of the parts of iCub robot (head, left_arm, right_arm, torso, left_leg and right_leg) are defined in terms of planning group, and for each of them a ros2_control of type FollowJointTrajectory is set.

ergocub_moveit_config

It contains the same information described in the previous paragraph, but customized with ergoCub specs.

xcub_moveit_test_controller

In this folder, a test to sample the reaching space is available. It can be run as a ros2 node with the provided launch file called test_controller.launch.py. For this purpose, you can find more info in the Use case paragraph above.

Moreover, a detailed report about the controller performance can be found here 📝.

Use case

Run the demos

As described in the previous section, this repository contains two demos that show the possibility to control a yarp-based robot within the MoveIt2 framework in the Cartesian space. First of all, make sure that yarpserver is running on your machine. Then, you can start launching a basic simulation:

# Build the packages within your ros2 workspace
cd ~/<ros2_ws>
source /opt/ros/humble/setup.bash 
colcon build
source install/setup.bash

# Remember to check if your robot name variable is set. If not:
export YARP_ROBOT_NAME="iCubGazeboV2_5"

# Launch the start-up nodes
ros2 launch xcub_moveit_robot robot_sim.launch.py

In this way, both rviz2 and gazebo windows are opened with the iCub model spawned in the two environments. At this point, open another shell, build and source the enviroment and then launch the ros2_control nodes:

cd ~/<ros2_ws>
source /opt/ros/humble/setup.bash 
colcon build
source install/setup.bash
export YARP_ROBOT_NAME="iCubGazeboV2_5"

# Launch the ros2_control nodes
ros2 launch xcub_moveit_robot robot_controls.launch.py

To verify that everthing has been done successfully, you can run in a separate shell:

ros2 control list_controllers

Here, you should see all the controllers loaded and their state (i.e. active or inactive).

Finally, open another shell and run the chosen demo, for example:

cd ~/<ros2_ws>
source /opt/ros/humble/setup.bash 
colcon build
source install/setup.bash
export YARP_ROBOT_NAME="iCubGazeboV2_5"

# For the grasping demo
ros2 launch xcub_moveit_robot grasp_demo.launch.py

If you want to see iCub performing a circle movement, instead of the last line, you can run:

ros2 launch xcub_moveit_robot circle_demo.launch.py

and follow the instructions on the third shell you opened. You should have something like this:

Controller performance

If you want to test the ros2_control xcub_ros2_controllers performance, you can rely to the xcub_moveit_test_controller package. It contains a simple script to sample the reaching space in front of iCub model. To run the simulation, please follow the first two steps described in the Run the demos section. After that, open another shell, source your workspace and then launch:

ros2 launch xcub_moveit_test_controller test_controller.launch.py

In this way, the test will start and the acquired data in terms of ideal and real poses are saved in two separated files in your current directory. Finally, you can plot them using the Matlab scripts provided in the utils folder.