Contact: [email protected]
ROS2 Grasp Library
A ROS2 intelligent visual grasp solution for advanced industrial usages, with OpenVINO™ grasp detection and MoveIt Grasp Planning.
Overview
ROS2 Grasp Library enables state-of-the-art CNN based deep learning grasp detection algorithms on ROS2 for intelligent visual grasp in industrial robot usage scenarios. This package provides ROS2 interfaces compliant with the open source MoveIt motion planning framework supported by most of the robot models in ROS industrial. This package delivers
- A ROS2 Grasp Planner providing grasp planning service, as an extensible capability of MoveIt (moveit_msgs::srv::GraspPlanning), translating grasp detection results into MoveIt Interfaces (moveit_msgs::msg::Grasp)
- A ROS2 Grasp Detctor abstracting interfaces for grasp detection results
- A ROS2 hand-eye calibration module generating transformation from camera frame to robot frame
- ROS2 example applications demonstrating how to use this ROS2 Grasp Library in advanced industrial usages for intelligent visual grasp
Grasp Detection Algorithms
Grasp detection back-end algorithms enabled by this ROS2 Grasp Library:
-
Grasp Pose Detection detects 6-DOF grasp poses for a 2-finger grasp (e.g. a parallel jaw gripper) in 3D point clouds from RGBD sensor or PCD file. The grasp detection was enabled with Intel® DLDT toolkit and Intel® OpenVINO™ toolkit.
Tutorials
Refer to ROS2 Grasp Library Tutorials for how to
- Install, build, and launch the ROS2 Grasp Planner and Detector
- Use launch options to customize in a new workspace
- Bring up the intelligent visual grasp solution on a new robot
- Do hand-eye calibration for a new camera setup
- Launch the example applications
Example Applications
Random Picking (OpenVINO Grasp Detection)
Recognition Picking (OpenVINO Grasp Detection + OpenVINO Mask-rcnn Object Segmentation)
Known Issues
- Cloud camera failed at "Invalid sizes when resizing a matrix or array" when dealing with XYZRGBA pointcloud from ROS2 Realsenes, tracked as #6 of gpg, patch under review.
- 'colcon test' sometimes failed with test suite "tgrasp_ros2", due to ROS2 service request failure issue (reported ros2 examples issue #228 and detailed discussed in ros2 demo issue #304)
- Rviz2 failed to receive Static TF from camera due to transient_local QoS (expected in the coming ROS2 Eloquent, discussed in geometry2 issue #183), workaround patch available till the adaption to Eloquent
Contribute to This Project
It's welcomed to contribute to this project. Here're some recommended practices:
- When adding a new feature it's expected to add tests covering the new functionalities
colcon test --packages-select <names_of_affected_packages>
- Before submitting a patch, it's recommended to pass all existing tests to avoid regression
colcon test --packages-select <names_of_existing_packages>