Home

Awesome

ReSwarm Distributed Nonlinear MPC

A package for Distributed Nonlinear MPC, based on CasADi, ACADOS and ROS middleware.

This package provides system models and Model Predictive Controllers to control a team of Astrobees free-flyers for both setpoint stabilization and trajectory tracking.

NOTE: This package requires an installed version of RTI DDS (along with the NASA-bundled RTI debians), and depends on the PR Simulation PMC Timeout Update.

Installation

This package depends on ROS Kinetic, but should be compatible with newer versions of ROS. It is assumed that the user previously installed ROS on his system.

  1. Install the Astrobee simulator following the instructions for General Users

  2. Clone reswarm_dmpc to your catkin_ws and build the workspace

cd catkin_ws/src
git clone https://github.com/Pedro-Roque/reswarm_dmpc.git
cd reswarm_dmpc/
pip install -r requirements.txt
cd ../../
catkin build
source devel/setup.sh

Testing

The ROS nodes are created with the *_interface.launch files. Here, we provide an example:

  1. Run the Astrobee simulator using the provided launch files
roslaunch reswarm_dmpc astrobee_sim.launch
  1. Run a unit test for translation with
roslaunch reswarm_dmpc unit_test_translation_interface.launch

or a formation control setting by launching on two terminals

roslaunch reswarm_dmpc leader_interface.launch

and

roslaunch reswarm_dmpc subleader_interface.launch
  1. Make sure that the robots are not in a faulty state by overriding it with
rostopic pub /honey/mgt/sys_monitor/state ff_msgs/FaultState '{state: 0}' & \
rostopic pub /bumble/mgt/sys_monitor/state ff_msgs/FaultState '{state: 0}'
  1. At this point, the node should show "Sleeping..." as a ROS Info message. This means that the node is waiting to be started. To start the node, call the starting service for each robot with
rosservice call /honey/start "data: true"

and

rosservice call /bumble/start "data: true"
  1. After a few seconds, the robots should be moving and the window where the unit test was launched will print information regarding the controller status, computational times, function costs, among other variables of interest.

Acknowledgements

A special thanks goes to Bryce Doerr, Keenan Albee, Monica Ekal and Rodrigo Ventura (the MPP crew), and to Brian Coltin and Rubén Ruiz, as well as to all the Astrobee Ops team, for their support in-view of the MPP ReSWARM test sessions and Astrobee Flight Software.

References

P. Roque, S. Heshmati Alamdari, A. Nikou, and D. V. Dimarogonas, "Decentralized Formation Control for Multiple Quadrotors under Unidirectional Communication Constraints", presented at the International Federation of Automatic Control (IFAC) World Congress, 2020. To cite this work:

@inproceedings{roque2020decentralized,
  title={Decentralized Formation Control for Multiple Quadrotors under Unidirectional Communication Constraints},
  author={Roque, Pedro and Heshmati Alamdari, Shahabodin and Nikou, Alexandros and Dimarogonas, Dimos V},
  booktitle={International Federation of Automatic Control (IFAC) World Congress},
  year={2020}
}