changhai0551/geograsp

Geometry-based method for computing grasping points on 3D point clouds

GeoGrasp

Geometry-based method for computing grasping points on 3D point clouds. Find more details at: https://www.researchgate.net/publication/331358070_Fast_Geometry-based_Computation_of_Grasping_Points_on_Three-dimensional_Point_Clouds

We are still working on improving the code efficiency and readability. Later, we will add tutorials on the use of GeoGrasp to work with robotic graspers.

Requirements

The package has been tested on Ubuntu 16.04. GeoGrasp is wrapped in a ROS package with the following dependecies:

• ROS Kinetic
• PCL 1.7

The rest of the dependencies (ROS packages) can be found at the package.xml file inside the GeoGrasp folder. In order to compile it, just clone this repository inside the source directory of your catkin workspace and execute catkin_make.

Examples of use

At GeoGrasp/data we have included two PCD files with two scenes. The creeper-isolated.pcd holds the 3D point cloud of a toy Creeper standing on a table. In contrast, objects-example.pcd contains a 3D point cloud in which there are multiple objets on a table. These clouds were captured with a Intel RealSense SR300 camera. To test GeoGrasp, simply execute the test script cloud_processor included in the repository:

rosrun geograsp cloud_processor _topic:="/cloud_pcd"

This launches a ROS node that subscribes to the topic /cloud_pcd in which point clouds will be published. The node reads these point clouds and processes them in order to compute the contact points using GeoGrasp. Next, execute the following node (pcl_ros ROS package is required for running this example):

rosrun pcl_ros pcd_to_pointcloud <file.pcd>

This publishes the contains of the PCD file <file.pcd> (either of those included at GeoGrasp/Data) on a topic called /cloud_pcd. See below an example of the computed points for the example PCD files:

Camera orientation

The camera is assummed to be oriented in such a way so its Z axe is pointing towards the table and the Y axe is pointing downwards, somewhat parallel to the table's normal vector. Check the two PCD files included to find an example of this point of view of the camera. In case of willing to use GeoGrasp with a camera on top of the table or another supporting surface (i.e. with its Z axe completely parallel to the table's normal vector), a minor change to the code base is required, so the initial grasping points are correctly found. The point2cloud branch has this change included in the file GeoGrasp.cpp, from line #131. We are working on a way of including this to the master branch so the type of configuration of the camera can be used as a parameter for changing this automatically.

Citation

[1] Zapata-Impata, B. S., Mateo, C. M., Gil, P., & Pomares, J. (2017). Using Geometry to Detect Grasping Points on 3D Unknown Point Cloud. In Proceedings of the 14th International Conference on Informatics in Control, Automation and Robotics (ICINCO) 2017 (Vol. 2, pp. 154–161). Best Paper Award. SCITEPRESS - Science and Technology Publications. https://doi.org/10.5220/0006470701540161

[2] Zapata-Impata, B. S., Gil, P., Pomares, J., & Torres, F. (2019). Fast geometry-based computation of grasping points on three-dimensional point clouds. International Journal of Advanced Robotic Systems, 16(1), 172988141983184. https://doi.org/10.1177/1729881419831846