This is a ROS package for GP-net to be used on mobile manipulators. It uses a GP-net model to propose up to 5 grasps based on a depth image. A pre-trained model for a robot with a PAL parallel jaw gripper is available at zenodo. If you want to use GP-net for alternative grippers, you have to train a new model by generating a new training dataset with the dataset generation code and training a new model with our code available on github.
This code has been tested with a ROS kinetic installation and python 2.7.
Besides the ROS installation, the package requirements are listed in requirements.txt.
We recommend an installation via docker or a virtual environment.
After cloning this package into the src folder in your catkin workspace, build the workspace
and source catkin_ws/devel/setup.bash in order to use the package.
GP-net can be used by launching the grasp_planning_node.py and use the planning service by
sending a depth image and the camera info to the node, e.g.:
plan_grasp = rospy.ServiceProxy('gpnet_grasp_planner', GPnetGraspPlanner)
depth_im = rospy.wait_for_message('/camera/depth_image', Image)
camera_intr = rospy.wait_for_message('/camera/info', CameraInfo)
grasp_response = plan_grasp(depth_im, camera_intr)
An example usage script is given in scripts/tiago_example.py, which can be used with
roslaunch gpnet tiago_example.launch
Note that you will have to adjust the model_dir in the launch file to the path were you
your GP-net model is stored. A pretrained model is available at zenodo
If you use this code, please cite
A. Konrad, J. McDonald and R. Villing, "GP-Net: Flexible Viewpoint Grasp Proposal," in 21st International Conference on Advanced Robotics (ICAR), (pp. 317-324), 2023.
This publication has emanated from research supported in part by Grants from Science Foundation Ireland under Grant numbers 18/CRT/6049 and 16/RI/3399. The opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Science Foundation Ireland.