DOPE implementation on AUBO robotic arm

[Rinvay67]

Now I have successfully deployed DOPE on the machine, and I can get the "/dope/pose_cracker" topic under the "camera_color_optical_frame". Also, I have successfully got the transform from the "base_link" to "camera_color_optical_frame", How can I get the transform of the object under the "base_link" ?

[TontonTremblay]

you have to run a static transform from your camera to your robot. If you have it then you can querry the poses. Does that make sense?

[edvart-ros]

Here is how I do it. Im not a ros expert but this seems pretty straight forward. DOPE already provides the pose of the object in the optical frame link, so all ih ave to do is subscribe to that and transform the broadcast.

#!/usr/bin/env python3

import roslib
import rospy
from geometry_msgs.msg import PoseStamped
import tf

def pose_callback(msg):
    position = msg.pose.position
    orientation = msg.pose.orientation

    br = tf.TransformBroadcaster()
    br.sendTransform((position.x, position.y, position.z),
                     (orientation.x, orientation.y, orientation.z, orientation.w),
                     rospy.Time.now(),"cheezit","camera_color_optical_frame")

if __name__ == '__main__':
    rospy.init_node('cheezit_tf_broadcaster')
    rospy.Subscriber('dope/pose_cracker', PoseStamped, pose_callback, queue_size = 10)
    rospy.spin()

[mintar]

It's better to use the detected_objects topic instead of the pose_* topics (like pose_cracker), especially if you have multiple objects of the same type. The poses inside the detected_objects message are in the optical frame, and you can create a tf2.TransformListener and call its transformPose method to transform them into an arbitrary tf frame, like base_link.

[edvart-ros]

It's better to use the detected_objects topic instead of the pose_* topics (like pose_cracker), especially if you have multiple objects of the same type. The poses inside the detected_objects message are in the optical frame, and you can create a tf2.TransformListener and call its transformPose method to transform them into an arbitrary tf frame, like base_link.

Awesome! I've installed this package just today, so this is great to know.