ArUco Markers

DelibFlow uses ArUco markers to give TiAGo a visual anchor point for object grasping. The tiago_aruco_broadcast package ties together marker detection and TF publishing so that behavior tree nodes can get the object’s pose directly from the TF tree.

Launch everything at once

The easiest way to get ArUco detection running is through the bringup launch file, which starts it automatically:

ros2 launch tiago_demos tiago_bringup.launch.py

If you need to run it standalone:

ros2 launch tiago_aruco_broadcast aruco.launch.py

There are two nodes involved, each doing a different job.

ArUco Recognition Node

This node reads the camera feed and detects markers in the image.

ros2 run ros2_aruco aruco_node

Topics subscribed

ROS Topic	Interface	Description
`/camera/image_raw`	sensor_msgs/msg/Image	Raw image from the robot’s camera
`/camera/camera_info`	sensor_msgs/msg/CameraInfo	Camera intrinsics and extrinsics (needed for pose estimation)

Topics published

ROS Topic	Interface	Description
`/aruco_markers`	ros2_aruco_interfaces/msg/ArucoMarkers	Array of detected marker poses with their IDs
`/aruco_poses`	geometry_msgs/msg/PoseArray	All detected poses in a single message, handy for RViz visualization

Parameters

Parameter	Default	Description
`marker_size`	`0.0625`	Physical size of the marker in meters
`aruco_dictionary_id`	`DICT_5X5_250`	ArUco dictionary used to generate the markers
`image_topic`	`/camera/image_raw`	Image topic to subscribe to
`camera_info_topic`	`/camera/camera_info`	Camera info topic to subscribe to
`camera_frame`	(empty)	Camera optical frame name

ArUco TF Broadcast Node

This node takes the detected marker poses and broadcasts them as TF transforms. This is what allows behavior tree nodes to look up the object’s position in the map or base_link frame.

ros2 run tiago_aruco_broadcast aruco_broadcast

Topics subscribed

ROS Topic	Interface	Description
`/aruco_markers`	ros2_aruco_interfaces/msg/ArucoMarkers	Detected marker poses coming from the recognition node

Topics published

ROS Topic	Interface	Description
`/tf`	tf2_msgs/msg/TFMessage	Transform from the camera frame to each detected marker

How it all fits together

The recognition node detects markers and publishes their poses. The broadcast node takes those poses and puts them into the TF tree. The behavior tree node tiago_set_target_using_aruco_marker then reads the TF transform for the target marker and sets the grasp goal on the blackboard. This is the pipeline that the pick_object_example demo uses.