Ros2_aruco

This repository contains the Vision Package for the first assignment of the Experimental Robotics Laboratory Course at the Robotics Engineering Master Course of Genoa. The authors are:

• Valentina Condorelli, 4945679;
• Annika Delucchi, 4975984;
• Ramona Ferrari, 4944096;
• Daniele Rialdi, 4964038.

The package was initially forked by the Professor's repository, to which we added the file vision_node.py at the path ros2_aruco/ros2_aruco/.

Assignment

The assignment required to:

• spawn the robot in Gazebo, surrounded by 5 Aruco markers arranged in a circle;
• implement a routine that starts finding the marker with the lowest ID, and then finds all markers in order;
• each time a marker is found, a new image is published on a custom topic with a circle around the marker found;
• implement the same behaviour with two different nodes: in the first one, the whole robot moves; in the second one, you find the markers by only moving the camera.

Note that to accomplish these tasks, the behaviour of this vision_node must be coupled with the control routine that can be found on this repository.

How to run it

Prerequisites

This package depends on a recent version of OpenCV Python bindings:

pip install opencv-contrib-python # or pip3

Commands

First of all, the simulation must be started with the following command, in a terminal:

ros2 launch robot_urdf gazebo_aruco.launch.py

Then, there are two types of nodes that need to be launched:

• control nodes (check the ros2 branch of our robot control repo)
• vision nodes

With the control nodes, there are two possibilities to make the robot move:

1. Make the robot rotate
2. Make the camera rotate

In the first case, the following command must be launched in a terminal:

ros2 run robot_urdf camera_controller_node

In the second case:

ros2 run robot_urdf robot_controller_node

In both cases, the robot can be stopped anytime by pressing q. The node will not be killed but just paused, so that it can be restarted by pressing r.

As for the vision part, two nodes need to be launched:

• aruco_node, part of the ros2_aruco package
• vision_node, the node we developed

Therefore, the following commands must be run in two different terminals:

ros2 run ros2_aruco aruco_node --ros-args --remap /image:=/robot/camera1/image_raw

and

ros2 run ros2_aruco vision_node

Already existing nodes

The implemented routine for detecting and sorting the markers relies on the behaviour of the already implemented aruco_node.py, which locates Aruco AR markers in images and publishes their IDs and poses.

Category	Name/Topic	Type/Description
Subscriptions	/camera/image_raw	sensor_msgs.msg.Image
	/camera/camera_info	sensor_msgs.msg.CameraInfo
Published Topics	/aruco_poses	geometry_msgs.msg.PoseArray - Poses of all detected markers (suitable for rviz visualization)
	/aruco_markers	ros2_aruco_interfaces.msg.ArucoMarkers - Array of all poses with corresponding marker ID
Parameters	marker_size	Size of the markers in meters
	aruco_dictionary_id	Dictionary used to generate markers
	image_topic	Image topic to subscribe to (/camera/image_raw)
	camera_info_topic	Camera info topic to subscribe to (/camera/camera_info)
	camera_frame	Camera optical frame to use

vision_node.py

This node, integrated with aruco_node.py, is designed to detect a sequence of ArUco markers in a specific order and manage their visualization and publication. It retrieves raw camera images and ArUco marker detections from other topics and processes them to track and display the detected markers.

Node Functionality

Type	Topic	Message Type	Description
Subscriptions	/camera/image_raw	sensor_msgs/Image	Captures the raw image feed from the camera.
	/aruco_markers	ros2_aruco_interfaces/ArucoMarkers	Receives the detected ArUco markers, including their IDs and poses, from aruco_node.py.
Publications	/aruco_markers/detected_marker_image	sensor_msgs/Image	Publishes an annotated image highlighting the detected ArUco marker to a custom topic for visualization.
Parameters	detected_marker_image_topic	-	Specifies the custom topic to publish the detected marker image (default: aruco_markers/detected_marker_image).

Behavior

At the very beginning, the node sets up a publisher for annotated images and subscribers for the raw camera feed and ArUco marker detections; then it initializes a CvBridge for ROS-OpenCV conversions and an image container for processing frames.

After the initialization process, the implemented logic follows these steps.

1. The robot starts rotating. During the rotation, it sees the Aruco Markers previously placed in the environment. The IDs of the markers are retrieved from the camera information and properly stored in a list of markers seen for the first time. Every time the robot sees a new marker, it suddenly publishes a picture of it on the custom topic detected_marker_image_topic.
2. The robot continues until all five markers have been detected at least once.
3. Once the markers have been seen at least one time, the list containing the IDs is sorted to plan how to publish their image in the correct order. To switch to this modality, in which the robot looks for a desired marker at a time, the boolean flag first_sight is used.
4. When the desired marker is found, a frame containing it is published one more time on the custom topic detected_marker_image_topic.
5. Once the sorted list is over, all the markers' images have been published again in the correct order, so the program ends.