View on GitHub

burg-toolkit

Toolkit for benchmarking and understanding robotic grasping

BURG-Toolkit: Reproducible Robot Grasping Experiments in Simulation and the Real World

Martin Rudorfer1, Markus Suchi2, Mohan Sridharan1, Markus Vincze2, AleŇ° Leonardis1

1School of Computer Science, University of Birmingham, UK
2Automation and Control Institue, TU Vienna, Austria

The BURG-Toolkit is a set of open-source tools for Benchmarking and Understanding Robotic Grasping. Our tools allow researchers to:

  1. create virtual scenes for generating training data and performing grasping in simulation,
  2. recreate the scene by arranging the corresponding objects accurately in the physical world for real robot experiments, supporting an analysis of the sim-to-real gap, and
  3. share the scenes with other researchers to foster comparability and reproducibility of experimental results.

Feel free to get in touch if you have any comments or questions!

Overview

The toolkit includes the following components:

An exemplary workflow using the BURG-Toolkit:

(a-c) After creating the virtual scene using the SetupTool, we can sample grasps and execute them in a physics-based simulation; (d) From the SetupTool we can export a printout with an ArUco marker board and the object height maps indicating the placements; (e) To further facilitate the placement, our SceneVisualiser projects the objects into the camera image for the user to make sure the real objects align visually; and (f) Finally, the scene is set up for real robot experiments.

BURG-Toolkit

At the core of the BURG-Toolkit is a Python package of the same name. The BURG-Toolkit serves as a backbone for the other components and offers the following features:

SetupTool

The SetupTool is a graphical front end of the BURG-Toolkit dedicated to virtual scene creation. It is implemented as a Blender add-on and handles the following tasks:

Scene composition

Scenes can be created in various sizes, the templates automatically adjust. If necessary, the templates can be split up for printing to the desired page size.

Printouts

SceneVisualizer

The SceneVisualizer is a ROS package for arranging the objects for real-world experiments. It uses Augmented Reality to project the objects into the real scene.

Setup of a Scene

Acknowledgments

This work was conducted within the BURG research project which is supported by CHIST-ERA under EPSRC grant no. EP/S032487/1 and FWF grant no. I3967-N30.