Robust Grasp Planning Over Uncertain Shape Completions

This repository includes code used in our work on robust grasp planning over uncertain shape completions. More specifically, all sub-modules used to run the simulated and real grasping experiment

Authors: Jens Lundell
Maintainer: Jens Lundell, [email protected]
Affiliation: Intelligent Robotics Lab, Aalto University

Getting Started

The code was developed for python2.7 and Ubuntu 18.04.

Dependencies

Graspit (You need to use our fork of Graspit! to make the code work.)
PyTorch
Curvox
binvox-rw-py
python-pcl
Mesh_Reconstruction
ROS Melodic

Sub modules

The code includes the following sub-modules that are automatically pulled in

• graspit_commander
• graspit_interface
• graspit_msgs
• grid_sample_plugin
• shape_completion_network
• point_cloud_segmentation
• Fast-3D-Pointcloud-Segmentation

ROS Installation

Clone or download the project from Github:

mkdir ws
cd ws
git clone --recursive [email protected]:aalto-intelligent-robotics/robust_grasp_planning_over_uncertain_shape_completions.git
catkin_make

Simulation experiments

First of all, you need to download the ground-truth meshes and create some shape completed meshes. Ground-truth meshes can be downloaded from here and for shape completed meshes you can either download some pre-generated ones here or generate new ones as detailed here.

After you have these meshes, open a terminal and run

roslaunch graspit_commander graspit_commander.launch

Then, open another terminal and run

roslaunch grid_sample_plugin grid_sample_plugin.launch

which will open the following user interface

Next, click on the Do Simulation Experiments button and the following window will pop up

Then, click each of these buttons and give the appropriate path to the data. Once you are done, simply click on run. If everything goes as expected, you will see the following in the original GraspIt! window

GraspIt! is now generating and evaluating the grasps in the background and once it is done, it will save the results in the folder you specified earlier.

Real-World experiments

Coming soon

Citation

This repository contains code corresponding to our work on robust grasp planning over uncertain shape completions. If you use the code for your research, please cite

@article{lundell2019robust,
  title={Robust Grasp Planning Over Uncertain Shape Completions},
  author={Lundell, Jens and Verdoja, Francesco and Kyrki, Ville},
  journal={arXiv preprint arXiv:1903.00645},
  year={2019}
}

License

This project is licensed under the MIT License - see the LICENSE file for details