WWU CSCI 497F / 597F Lab 2

Goals

1. Become familiar with the basic functionality of ROS2 and Gazebo, an open source 3D physics simulator.

2. Become familiar with sensor modelling and feedback system design.

Deliverables

You are expected to upload a typed lab report. Please fork this repository and include a link to your forked version in your lab report.

For each numbered item in the lab assignment, the report should contain:

• Methods: a summary of what you did in this lab. Should be written in your own words, can be short but needs to include enough details so that I understand what decisions you made and why.
• Results: a summary of what you learned, in your own words.

At the end of your lab report, include:

• Next Steps: a list of questions that you have after completing this lab, and/or what you would do next to fix your system if it’s not working.

Usage

1. Build the project

   colcon build --cmake-args -DBUILD_TESTING=ON

2. Source the workspace

   source ./install/setup.sh

3. Launch the simulation

   ros2 launch ros_gz_bringup diff_drive.launch.py

For a more detailed guide on using this package see the Gazebo documentation.

Assignment

1. Launch the simulation. Two windows should launch: the Gazebo simulation of a differential drive robot in a maze and RViz, a ROS program that visualizes topics such as pose and odometry estimates and sensor readings.

   • Note: Each time you launch the simulation, to you need to press the arrow-shaped "Play" button in the lower left hand corner of the Gazebo simulator.
   • In a new terminal with ROS2 configured, run the command

   ros2 topic pub /diff_drive/cmd_vel geometry_msgs/msg/Twist "{linear: {x: 5.0, y: 0.0, z: 0.0}, angular: {x: 0.0, y: 0.0, z: -0.1}}"

   • In another terminal with ROS2 configured, run the command

   ros2 topic echo /diff_drive/scan

   What do you notice about the output of the second command as the robot moves around the simulated space? This topic publishes the current distance measured by a single-laser range detector.

2. Using publisher and subscriber nodes, implement a wall-following strategy that uses the laser scan data to ensure the robot can follow the wall all the way around the environment. Submit a screenshot of the RViz window with the odometry path of your robot in your writeup. The figure below shows an example of a high-level strategy for wall-following. The numbers are arbitrary and you may structure your controller differently. For example, you may design a continuous controller instead of discrete.

Notably, our robot lacks a side-facing sensor in the original version. You may choose to turn the robot (while in motion, or in place) to simulate such a measurement, or you may choose to edit the sensor model to allow for a wider "sweep" of the laser rangefinder. You can find our sensor model in ros_gz_gazebo/worlds/diff_drive.sdf (search for "lidar"). The documentation for the different description fields for the lidar sensor can be found here.

For an A grade, your robot only needs to be able to circumnavigate the larger room; you will be graded based upon the level of understanding in your writeup. If your robot misses a turn or gets stuck, try to explain why you think it is happening.

3. Use CTRL+C to close the simulator and RViz. Open the file ros_gz_description/models/diff_drive/model.sdf in a text editor. Search for the word "sensor" to find the definition of the laser scanner.

   • a. In the <range> block, change the max range to be 1.0 instead of 100.0 . Save the file, rebuild the entire lab2 ROS package, and restart your simulation. What is the effect on your wall-following strategy? How can you change your strategy to compensate for the reduced sensing range? Reset your system to have a maximum range of 100.0 when you are done.

   • b. In the <noise> block, change the standard deviation of the sensor noise (stddev) to be 1.0 instead of 0.1. Save, rebuild, and restart your simulation. What is the effect on your wall-following strategy? How can you compensate for the large noise level? Reset your system to a noise level of 0.1 when you are done.

   • c. In the <noise> block, change the mean to be nonzero. What is the effect on your wall-following strategy? What happens as the value of the mean gets larger? What happens as the mean gets smaller (the sensor experiences a stronger negative bias)?

4. (Graduate students only) Implement a low-pass filter on your rangefinder data in your wall-following code. Analyze the performance of your wall-following strategy with respect to both:

   • a. Varying the standard deviation of the rangefinder noise; and
   • b. Varying the baseline speed of the robot. Ideally, include a plot each for a) and b), graphing a metric of the performance of your system over time. Screenshots of RViz are OK too, as well as precise technical descriptions of the resulting robot motion.

Recommended Reading

• Chapter 2, through section 2.4: Sensing for a Point Robot of Computational Principles of Mobile Robotics (CPMR).
• Chapter 3.1 of CPMR, specifically the parts on differential drive robots.
• Chapter sections 11.1, 11.2, and 11.6 (for graduate students) of Planning Algorithms.

Included packages

• ros_gz_description - holds the sdf description of the simulated system and any other assets.

• ros_gz_gazebo - holds gazebo specific code and configurations.

• ros_gz_app - holds ros2 specific code and configurations.

• ros_gz_bringup - holds launch files and high level utilities.