DetectGraspablePosesPcabase¶

What Is This¶

../../_images/detect_graspable_poses_pcabase.png

Detect_graspable_poses_pcabase.py is a program which publishes pose array where a robot can grasp, using point cloud of a target object and PCA algorithm. In the above image, the grasp poses are visualized with the axies. Please transform input cloud to a flame which direction is the same as the robot (e.g. /BODY) before using this program.

../../_images/detect_graspable_poses_pcabase.gif

The above image illustrates an example way to use this node. Axies represent poses where the robot can grasp.

Example videos using a real robot (only allowed for the jsk members)¶

Subscribing Topic¶

~input (sensor_msgs/PointCloud2)

Input pointcloud. RGB field is required.

Publishing Topic¶

~output/can_grasp_poses (geometry_msgs/PoseArray)

PoseArray where a robot can grasp.

Parameters¶

~direction (Character, default: x)

From which direction a robot try to grasp a target object. (’x’: from the front, ‘z’: from the top)
~hand_width[m] (Float, default: 0.13)

How long a robot can spread its hand.
~interval_m[m] (Float, default: 0.04)

Interval between target poses. Please decrease this number if you want to detect more possible grasp poses.