/wheelchair-ds-motion

DS-based motion planning for the quickie-salsa wheelchair simulated in Gazebo.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

wheelchair-ds-motion

DS-based motion planning for the quickie-salsa wheelchair simulated in Gazebo, as shown below:

Dependencies

To run this package you must install the following dependencies:

  • quickie-salsa-m2 checkout 'nadia' branch | Control Interface for Quickie-Salsa Wheelchair in Gazebo
  • ds-motion-generator checkout 'nadia' branch | DS motion generation nodes
  • lpvDS-lib | lpv-DS class used by ds-motion-generator, should be installed automatically if using wstool with the ds-motion-generator package.

Instructions

Step 1 Bring-up Gazebo Wheelchair Simulator of the Road World and RViz for DS Visualization

$ roslaunch wheelchair_ds_motion ds_simulation.launch world:=_road

Step 2 To run a non-linear DS (lpv formulation) with streamline visualization in rviz:

  • Load the DS model

     $ roslaunch wheelchair_ds_motion run_nonlinearDS_controller.launch

    The attractor and type of DS are set in "DS_name" parameter, there are currently 2 options:

     <arg name="DS_name" value="2D-W-Nav"/>
 <arg name="DS_name" value="2D-U-Nav"/>

    which points to the .yml file in the ds-motion-generator package.

  • To control the wheelchair with the loaded DS, run the following command:

     $ rosrun wheelchair_ds_motion nonlinearDS_controller.py

To learn your own lpv-DS models, download and follow the instructions in the ds-opt package.

Optional To run a simple linear DS with a pre-defined attractor:

$ roslaunch wheelchair_ds_motion run_linearDS_controller.launch

To define the attractor and if obstacles should be present or not, modify the following line:

<arg name="ctrl_command"   default="10 0 1" />
  • parameters: <x-position of attractor> <y-position of attractor> <number of obstacles>

Without obstacle avoidance, simply set the last parameter to 0.

References

[1] Figueroa, N. and Billard, A. (2018) "A Physically-Consistent Bayesian Non-Parametric Mixture Model for Dynamical System Learning". Conference on Robot Learning (CoRL) - 2018 Edition. To Appear.

Contact: Nadia Figueroa (nadia.figueroafernandez AT epfl dot ch)

Acknowledgments This work was supported by the EU project Cogimon H2020-ICT-23-2014 and Crowdbot H2020-ICT-25-2016-2017.