ROS 2 Training Exercises

Yoan Mollard, for Nobleprog

Switch back to course

Exercises for DAY 1

1. Prepare Your Work Environment
2. turtlesim, CLI Tools
3. turtlesim, Client Libs
4. RMW Bridge to TCP/IP, MQTT or WS [uncorrected]

Exercises for DAY 2

5. Discover Tiago's ROS 2 Integration
6. Tiago 2D Navigation
7. Demonstrate micro-ROS [uncorrected]

Exercises for DAY 3

7. Tiago Manipulation with the Arm / Gripper
8. Motoman hc10dtp/BFFP Manipulation [uncorrected]
9. Demonstrate MoveIt Servo [uncorrected]

Exercise 1: Prepare Your Work Environment

You need to install ROS 2 Jazzy:

• Option A: With a container (preferred option for the training)
• Option B: Install ROS natively on Linux, Windows or MacOS
• Option C: Install ROS in a VM

Options B/C: Pick an OS that is Tier 1 for your ROS version (tier 1 = fully supported)

Option C: Make sure that 3D hardware acceleration is enabled. Not-that-great option for powerful Gazebo simulations, also because of network management.

The Jazzy Container setup for the course

Pre-installed devices:

• PAL Tiago, but in alpha version (FT sensor broken + some bugs)
• Motoman hc10dtp
• BFFP

Download the container:

git clone https://github.com/cboostpverwimp/ros2-course

Install Visual Studio Code and start the container:

code ros2-course/

Upon startup, click Trust the authors, accept the installation of the Dev Container extension, and reopen the folder in the container. Read the README.

Exercise 2: turtlesim, CLI Tools

The turtlesim Robot

• 2D "visual" turtle robot
• Simulates a wheeled robot (Turtlebot, Husky, etc.)
• Velocity control (r, θ):
  • Linear velocity r (in m/s)
  • Angular velocity θ (in rad/s)
• Stops automatically without receiving a new command

Follow the ROS tutorials in the Tutorials > Beginner: CLI tools category.

https://docs.ros.org/en/jazzy/Tutorials/Beginner-CLI-Tools.html

Exercise 3: turtlesim, Client Libs

Follow the tutorials in Tutorials > Beginner: Client libraries.

https://docs.ros.org/en/jazzy/Tutorials/Beginner-Client-Libraries.html

Follow at least purple tutorials

# If RAM is saturated during compilation:
sudo dd if=/dev/zero of=/swapfile bs=1M count=8000
sudo mkswap ~/swapfile
sudo swapon ~/swapfile
# compile... then disable swap
sudo swapoff ~/swapfile
rm ~/swapfile

Exercise 4: RMW Bridge to TCP/IP, MQTT or WS

Setup a PoC that demonstrates a bidirectionnal communication between the RMW and a MQTT or TCP/IP or WebSocket device.

You can cover: topics, service calls, action services (e.g. Nav2 goal)

Example A: IMU sensor connected to a PLC, streaming both:

• a /imu topic of type sensor_msgs/Imu
• a TF frame, viewable in RViz (use a TF Broadcaster)

Make it a rclpy.lifecycle.LifecycleNode so that it can be started/stopped.

Example B: Motor drive connected to a PLC (or Gazebo), commanded via a HTML GUI through WebSockets (see rosbridge for ROS 2)

Exercise 5: Discover Tiago's ROS 2 Integration

In all parts, refer to the
summary of commands from the container README

First start the Gazebo simulation of Tiago + teleoperation and observe the RTF percentage at the bottom right:

Real Time Factor (RTF): RTF indicates how fast Gazebo can simulate the scene. It is configured to be at 100%.

Below 100%, the simulation is slowed down, because of performance issues. Below 50%, it might be unusable.

Part 1: Monitor Joint States

With a ROS 2 CLI command, monitor the /joint_states topic representing the current state of all the robot's joints: their angular position (rad or m), angular velocity (rad/s or m/s), and effort (Nm).

How many joints do you find on this robot?

Identify the names of the 2 joints among these that do not have torque sensors. You will recognize them because they return the value NaN = Not A Number.

Using ros2 topic hz <topic>, determine the frequency (in Hertz) at which the robot publishes the state of its joints on this topic.

Depending the situation that frequency might be too low (e.g. for closed-loop motor control) but it is usually configurable in the ROS driver.

Part 2: Plot Joint States: Second Head Motor

Start the graphical trajectory controller: this allows individual control of each joint:

ros2 run rqt_joint_trajectory_controller rqt_joint_trajectory_controller

Try different motor controllers available in the list. Activate each controller with the red button and move the angular commands of the joints. Verify that your robot moves in Gazebo.

To continue, choose the head controller: it has two joints.

We will focus on head_2_joint.

Plotjuggler is a tool that plots curves in real-time (better than rqt_plot).

Start Plotjuggler: ros2 run plotjuggler plotjuggler

On the left in Streaming:

• Set the buffer to 120 seconds
• Click Start for ROS 2 Topic Subscriber
• Choose /joint_states
• Expand joint_states in the list on the left and open the head_2_joint
• Drag its torque value to the graph on the right to start

The graph fluctuates a lot, but observe the scale: it's just noise.

Move head_2_joint with the graphical controller and observe the effort graph in real-time.

Part 3: Monitor Sensors with RViz

Start RViz2: ros2 run rviz2 rviz2

Click Add to add new Display components to visualize and configure them by expanding their config

For each of the topics on the next page, use both the CLI (ros2 topic echo) and RViz2 (if possible) to visualize their data.

Teleoperate the robot to place it in front of or away from obstacles to observe differences in the data.

Make sure neither RViz2 nor the Display is in error

• LIDAR data (scan_raw)
• Inertial measurement unit data (/imu_sensor_broadcaster/imu)
• Force sensor (ft_sensor_controller/wrench) (broken in Jazzy / working on it)
• Sonar data (sonar_base) (this sonar simulation works weirdly in Jazzy)
• The robot's transformation tree with a TF type display.
  In the tree, identify where the robot's IMU is located by checking only the frame (= reference frame) corresponding to its attachment (base_imu_link).
  Which axis x, y, or z points forward on the robot?
• Odometry /mobile_base_controller/odom:
  • In the CLI, extract the position from each message by selecting a field: --field pose.pose.position
  • In RViz2, change the global frame to that of the messages (odom)
• The robot's URDF model (robot_description, via a RobotModel display)

Finally, with an Image display, also visualize the raw uncompressed RGB-D camera images in RViz2:

• image_raw for the raw image
• rgb for the RGB image
• depth_registered for depth (registered means the image is realigned for pixel-to-pixel correspondence with the RGB image)

If the depth image appears striped, it is a RViz2 bug. Use another ROS tool to visualize it:

ros2 run image_tools showimage image:=/head_front_camera/depth_image

RViz2 cannot decompress compressed images itself. If needed, use ros2 run image_transport republish for decompression.

Part 4 (if needed): Modify SDF worlds/models in Gazebo

Simulated worlds and models (objects) are SDF files that you can manually edit.

Some models are static: they are not subject to forces, not even gravity. To change this, modify <static>false</static> in the SDF.

A common place to download models is the
Gazebo Fuel server: https://app.gazebosim.org/

Each model can also be manually modified:
sdf_worlds/#download-the-model

There are also many opensource pre-designed worlds, eg https://github.com/aws-robotics/ (take the Jazzy or ROS 2 branch. Complex worlds will be slow)

Remember to compile and source after modifying a world or model.

Exercise 6: Tiago 2D Navigation

Part 1: Teleoperated Navigation

Consult the container README to obtain the commands to teleoperate the robot's mobile base using the keyboard.

Open rqt_graph as seen in previous tutorials and note the hierarchical name (with slashes) of the velocity command topic for the mobile base.

Observe the 2 other nodes present between the teleoperation node and the mobile base controller: a multiplexer (mux) + a stamper (to sync time).

Part 2: Autonomous Navigation

Consult the container README to obtain the commands to:

1. Map the simulated scene using SLAM according to the documentation.

2. Save the map to a file named our_map and CLOSE SLAM. Compile if needed.

3. Start tiago_nav2d to use the saved map by giving a goal in RViz.

4. Move the robot to a target position using teleoperation and retrieve its coordinates on the map from the /amcl_pose topic.

5. Develop an rclpy node in a new package to reach the target at the recorded coordinates (Drafts C++ and Python).

6. Set up an infinite patrol of the robot between 2 points.

7. Insert an obstacle in Gazebo in the middle of the path: it should be avoided.

Exercise 7: Demonstrate micro-ROS

1. Flash the microcontroller with the microROS firmware
2. On the host machine, create a ROS 2 node that will act as the micro-ROS agent
3. Retrieve data from/to the microcontroller e.g.:

• /battery_status of type sensor_msgs.msg.BatteryState
• /imu of type sensor_msgs.msg.Imu

Rely on that video demo

Exercise 8: Tiago Manipulation with Arm/Gripper

Part 1: Kinematics

With the help of the container README, start MoveIt2 to display the colored manipulation handles in RViz.

Observe the live IK solutions obtained by moving the blue ball of the effector.

With the blue ball, define a target that does not put the robot in collision, then click Plan & Execute to execute the movement to the target.

In the Joints tab, move the Nullspace exploration slider to explore the Nullspace of the current end effector pose.

Part 2: Planning

Three robots of different colors overlap in RViz:

• Real colors (black and white)
• Orange
• Transparent (only briefly visible with each click on Plan & Execute)

What is the difference between these three?

To answer, use the Query X checkboxes in the Planning Request dropdown list in the RViz displays.

Part 3: Collision Avoidance

Open the Context tab, observe the list of planners available in the OMPL Library (RRT, PRM, EST, ...), and test different goals and plannings.

In the RViz displays, uncheck all necessary boxes so that your robot no longer appears at all. Then check only these: MotionPlanning Scene Robot Show Robot Collision.

Observe the collision model in grayscale, this is the 3D view of the URDF collision model.

Restart RViz to reset the displays.

In the Scene Objects tab, insert a Box and move it in front of the robot so that it is not in collision but obstructs its passage.

Define a target and click Plan & Execute. Verify that the object is avoided.

There seems to be a bug in this version of RViz that makes manipulating the scene objects difficult, or even causes it to crash. If it is unusable, skip to the next part.

Scene objects can also be inserted via Python or come from a RGB-D camera (voxels).

Part 4: Other Joint Groups of Tiago

In RViz, the Planning group list includes not only the arm but also arm_torso and gripper.

Try different Plan & Execute targets with the kinematic chain arm_torso.

Test also the arm group to observe the difference.

Finally, try with gripper. The gripper is not a kinematic chain, so it has no colored handles. Use Goal State: random valid to open and close the gripper*.

* Not very handy but fortunately this can be commanded from Python. Also, static poses could have been saved to the SRDF so that they can be clicked in MoveIt (e.g. open_gripper)

Part 5: Automate the Arm

We will control the arm with Python code. A draft node is already coded in the tiago_pick_and_place package.

With the help of the container README, open the node's code and execute it using its launchfile plan.launch.py.

The node has 3 functions corresponding to 3 different trajectory plans. For each of them, uncomment its function call and test it:

• Plan 1: Define a target in joint space.
• Plan 2: Define a target in Cartesian space.
• Plan 3: Open and close the gripper.

Exercise 9: Motoman hc10dtp/BFFP Manipulation

Open project: define your own goals

Exercise 10: Demonstrate MoveIt Servo

Open project: define your own goals

MoveIt Servo facilitates realtime control of your robot arm. MoveIt Servo accepts any of the following types of commands:

• Individual joint velocities
• The desired velocity of end effector
• The desired pose of end effector

Try to demonstrate MoveIt Servo on Motoman hc10dtp/BFFP.

See MoveIt Servo doc