SCA_data_construction

This repository includes the packages and instructions to generate a data set for Self-Collision Avoidance (SCA) between two or more arms. You can find the paper here:

and the corresponding video here:

This package does not learn the SCA boundary. It just generates the SCA data-set.

Dependencies

Mathlib form Robot-toolkit (https://github.com/epfl-lasa/robot-toolkit)

KUKA Rviz visualization (https://github.com/epfl-lasa/kuka-rviz-simulation) (Actually, you don't need it if you only want to generate the SCA data-set. This package is used to vitualize the motion of the robots in plot_on_robot.cpp)

KUKA FRI bridge (https://github.com/nbfigueroa/kuka_interface_packages) (Actually, you don't need it if you only want to generate the SCA data-set. This package is used to vitualize the motion of the robots in plot_on_robot.cpp)

Mlpack (https://github.com/mlpack/mlpack) (Actually, you don't need it if you only want to generate the SCA data-set. This package is used to construct a probabilistic model for the reachable workspace of each robot. I prefer Matlab for doing this. The matlab code is also included in this package.)

Features:

• Generating the data for self-collision boundary for two or more arms.

• Vitalizing the robot configurations on the Rviz simulator.

• Generating the data set of the positions of one end-effector.

Before make!

1. Open common.h. and change folder_path to your home folder.
2. To change the resolution of sampling and the position of the bases of the robots, open constructing_data_set.cpp.
3. To modify the kinematic of the robots or change the constraints on the joint workspaces, open constructing_data_set.cpp.

How to run

1.Generating SCA data-set

Make the data-set of each robot :

rosrun  constructing_data_set constructing_data_set

Then analysis the data set and find the collided configuration and boundaries of the collided configurations :

rosrun  constructing_data_set analysing_data_set

1.1 Visualizing the generated data set

Launch Rviz simulator with the correct parameters of each robot, for more information see KUKA Rviz visualization.

roslaunch kuka_lwr_bringup bimanual_simulation.launch

Run

rosrun  constructing_data_set plot_on_robot

2. Constructing a model of the reachbale space of a robot

Open Learning_the_workspaces.cpp and edit it accordingly. Then make the package and run

If you want to learn the data set in C++, you need to uncomment some lines in this file.

rosrun  constructing_data_set Learning_the_workspaces

Open Matlab and run Learning_Workspace.m

Copyright

Please cite these papers if you are using this toolbox: @article{mirrazavi2018unified, title={A unified framework for coordinated multi-arm motion planning}, author={Mirrazavi Salehian, Seyed Sina and Figueroa, Nadia and Billard, Aude}, journal={The International Journal of Robotics Research}, pages={0278364918765952}, publisher={SAGE Publications Sage UK: London, England} }

For more information contact Sina Mirrazavi.