/EDLUT_BAXTER

Cerebellar-SNN control of a Baxter robot. The repository includes EDLUT simulator source code, the ROS package to perform a closed-loop cerebellar-SNN control and all configuration files needed.

Primary LanguageC++

EDLUT_BAXTER

Please, cite as:

Abadía, I., Naveros, F., Garrido, J. A., Ros, E., & Luque, N. R. (2019). On robot compliance: a cerebellar control approach. IEEE transactions on cybernetics, 51(5), 2476-2489.

Cerebellar-SNN control of a Baxter robot. The repository includes EDLUT simulator source code, the ROS package to perform closed-loop cerebellar-SNN control and all configuration files needed.

There is also a folder containing the results obtained for each of the performed motor tasks. For each of the performed trajectories the following files are included:

  • recorded_desired_positions.txt : This file contains the desired position of each joint at each point of time along the experimental setup performed.
  • recorded_current_positions.txt : This file contains the current robot's position for each joint at each point of time along the experimental setup performed.
  • recorded_joint_MAE.txt : This file contains the Poisition Mean Absolute Error (MAE) (i.e. difference between desired and current position) of each joint for each of the trials performed along the experimental setup.
  • recorded_global_MAE.txt : This file contains the global (i.e. mean of all joints) Poisition MAE of each trial performed along the experimental setup.

Requirements

  • A computer with at least 8GB of RAM, a multicore CPU, an NVIDIA GPU with CUDA support, and Linux (Ubuntu 15.10 or Ubuntu 16.04, required for ROS Kinetic).
  • A Baxter robot (it can be simulated using Gazebo, although its behaviour is completely different from the real one).

Installation

  • Install and configure ROS (Kinetic distribution) and setup Baxter robot using the SDK Baxter guide: http://sdk.rethinkrobotics.com/wiki/Getting_Started

  • Install Gazebo (OPTIONAL): http://sdk.rethinkrobotics.com/wiki/Baxter_Simulator . Notice that the dynamics model of the simulated version is completely different from the real robot, obtaining different results.

  • Download source code from repository https://github.com/EduardoRosLab/EDLUT_BAXTER and copy in home folder.

  • Install EDLUT simulator (this step requires an NVIDIA GPU with CUDA support and CUDA installation):

    • Open a terminal and go to the folder EDLUT_source_code inside EDLUT_BAXTER repository.
    • chmod u+x configure
    • ./configure
    • make
    • sudo bash
    • make install

  • Compile the ROS EDLUT-BAXTER package (this step requires the installation of ROS and proper Baxter configuration):

  • Copy the config_files folder from EDLUT_BAXTER repository to ~/.ros path.

Execution

Using real Baxter robot

  • Connect to Baxter robot following http://sdk.rethinkrobotics.com/wiki/Hello_Baxter
    • Open a terminal:
    • cd ros_ws/
    • ./baxter.sh
    • rostopic pub /robot/joint_state_publish_rate std_msgs/UInt16 500 (Set Baxter publishing rate to 500 Hz)
    • cd ..
    • cd EDLUT_BAXTER/edlut_ros_ws
    • source devel/setup.bash
    • Launch the desired motor task:
      • roslaunch edlut_ros circle_trajectory.launch (Circle trajectory – Cerebellar-SNN Torque Control)
      • roslaunch edlut_ros eight_trajectory.launch (Eight-like trajectory – Cerebellar-SNN Torque Control)
      • roslaunch edlut_ros target_reaching.launch (Target Reaching – Cerebellar-SNN Torque Control)
      • roslaunch edlut_ros position_control_mode_circle_trajectory.launch (Circle trajectory – Position Control Mode)
      • roslaunch edlut_ros position_control_mode_eight_trajectory.launch (Eight-like trajectory – Position Control Mode)
      • roslaunch edlut_ros position_control_mode_target_reaching.launch (Target Reaching – Position Control Mode)
  • To use the monitoring tools run the rqt_reconfigure ROS plugin:
    • Open a terminal:
    • cd ros_ws/
    • ./baxter.sh
    • rosrun rqt_reconfigure rqt_reconfigure (A new window will be displayed allowing the user to choose among several monitoring tools)

Using simulated Baxter robot on Gazebo

  • Launch Baxter on Gazebo

    • Open a terminal:
    • cd ros_ws/
    • ./baxter.sh sim
    • roslaunch baxter_gazebo baxter_world.launch

  • Connect to simulated Baxter robot

    • Open a new terminal:
    • cd ros_ws/
    • ./baxter.sh sim
    • rostopic pub /robot/joint_state_publish_rate std_msgs/UInt16 500 (Set Baxter publishing rate to 500 Hz)
    • cd ..
    • cd EDLUT_BAXTER/edlut_ros_ws
    • source devel/setup.bash
    • Launch the desired motor task:
      • roslaunch edlut_ros circle_trajectory.launch (Circle trajectory – Cerebellar-SNN Torque Control)
      • roslaunch edlut_ros eight_trajectory.launch (Eight-like trajectory – Cerebellar-SNN Torque Control)
      • roslaunch edlut_ros target_reaching.launch (Target Reaching – Cerebellar-SNN Torque Control)
      • roslaunch edlut_ros position_control_mode_circle_trajectory.launch (Circle trajectory – Position Control Mode)
      • roslaunch edlut_ros position_control_mode_eight_trajectory.launch (Eight-like trajectory – Position Control Mode)
      • roslaunch edlut_ros position_control_mode_target_reaching.launch (Target Reaching – Position Control Mode)

  • To use the monitoring tools run the rqt_reconfigure ROS plugin:

    • Open a terminal:
    • cd ros_ws/
    • ./baxter.sh sim
    • rosrun rqt_reconfigure rqt_reconfigure (A new window will be displayed allowing the user to choose among several monitoring tools)