Cartesian Controllers

This package provides a set of Cartesian motion, force and compliance controllers for the ros2_control framework. The controllers are meant for joint position and joint velocity interfaces on the manipulators. As a unique selling point, they use fast forward dynamics simulations of virtually conditioned twins of the real setup as a solver for the inverse kinematics problem. Integrating from joint accelerations to joint velocities and joint positions gives them a delay-free, noise suppressing, and an inherently more stable contact behavior than conventional Admittance controllers. The controllers from this package are designed to trade smooth and stable behavior for accuracy where appropriate, and behave physically plausible for targets outside the robots reach. The package is for users who require interfaces to direct task space control without the need for collision checking.

The ROS1 version is here. Also see this talk at ROSCon'19 and these slides to get a brief overview.

Why this package?

Users may refer to MoveIt for end-effector motion planning, but integrating a full planning stack is often unnecessary for simple applications. Additionally, there are a lot of use cases where direct control in task space is mandatory: dynamic following of target poses, such as visual servoing, teleoperation, Cartesian teaching, or any form of closed loop control with external sensors for physical interactions with environments, such as Machine Learning applications. This package provides such a controller suite for the ros2_control framework.

Installation

Switch into the src folder of your current ROS2 workspace and

git clone -b ros2 https://github.com/fzi-forschungszentrum-informatik/cartesian_controllers.git
rosdep install --from-paths ./ --ignore-src -y
cd ..
colcon build --packages-skip cartesian_controller_simulation cartesian_controller_tests --cmake-args -DCMAKE_BUILD_TYPE=Release

This builds the cartesian_controllers without its simulation environment. The simulation is mostly relevant if you are just getting to know the cartesian_controllers and want to inspect how things work. You can install it according to this readme.

Now source your workspace again and you are ready to go.

Getting started

This assumes you have the cartesian_controller_simulation package installed. In a sourced terminal, call

ros2 launch cartesian_controller_simulation simulation.launch.py

This will start a simulated world in which you can inspect and try things. Here are some quick tutorials with further details:

• Solver details
• Cartesian motion controller
• Cartesian force controller
• Cartesian compliance controller
• Cartesian controller handles
• Teleoperation
• Example on Universal Robots

Citation and further reading

If you use the cartesian_controllers in your research projects, please consider citing our initial idea of the forward dynamics-based control approach (Paper):

@InProceedings{FDCC,
  Title                    = {Forward Dynamics Compliance Control (FDCC): A new approach to cartesian compliance for robotic manipulators},
  Author                   = {S. Scherzinger and A. Roennau and R. Dillmann},
  Booktitle                = {IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  Year                     = {2017},
  Pages                    = {4568-4575},
  Doi                      = {10.1109/IROS.2017.8206325}
}

If you are interested in more details, have a look at

• Inverse Kinematics with Forward Dynamics Solvers for Sampled Motion Tracking (Paper)
• Virtual Forward Dynamics Models for Cartesian Robot Control (Paper)
• Contact Skill Imitation Learning for Robot-Independent Assembly Programming (Paper)
• Human-Inspired Compliant Controllers for Robotic Assembly (PhD Thesis, especially Chapter 4)

Here's an application of imitation learning for force-controlled assembly on a UR10e

• Learning Human-Inspired Force Strategies for Robotic Assembly (Paper)