ModernCS
Join for free
ROS

Frontier · Taught with ROS

Robotics

ROS 2, control loops, mapping and navigation. You drive a simulated robot, make it find its own way across a mapped space, and learn why the physical one behaves differently.

Download the course plan
  • Foundational
  • 5 phases
  • 25 sessions
  • First session free

Syllabus

5 phases · 25 sessions · each phase ends in something you have built

1

Before Anything Moves

Get a Linux machine, a shell, and enough Python to write a loop that decides something.

$58

this phase

  1. 1.1

    What a Robot Actually Is

    Free preview

    Sense, decide, act, and why every robot you build is that loop running fast.

    Open
  2. 1.2

    Ubuntu the Way ROS Expects It

    Install Ubuntu 24.04 in a VM or on metal, and get around it from the terminal alone.

  3. 1.3

    Python That Does Not Stop

    Variables, functions, loops, and a program that keeps running until you kill it.

  4. 1.4

    A Controller Before ROS

    Error, gain, and a proportional controller you write with no framework underneath it.

By the end of this phase:An Ubuntu 24.04 machine you set up yourself, running a proportional controller you wrote in plain Python.
2

The Graph From the Command Line

Inspect and drive a running robot without writing a line of code.

$58

this phase

  1. 2.1

    Sourcing, and Why Nothing Works Yet

    Source the setup file, set ROS_DOMAIN_ID, and fix the two failures that stop most beginners.

  2. 2.2

    Nodes and Turtlesim

    Start nodes with ros2 run, list what is running, and move a robot with ros2 topic pub.

  3. 2.3

    Topics, Services and Actions

    The three ways nodes talk, told apart by watching each one live from the CLI.

  4. 2.4

    Parameters and rqt

    Retune a running node without restarting it, and read its logs in rqt_console.

  5. 2.5

    Launch Files and Bags

    Start a whole system with one command, record everything it said, and play it back.

By the end of this phase:A recorded bag of a turtlesim run, replayed, with a written inventory of every node, topic and service in it.
3

Code You Wrote

Build your own packages and close a control loop inside a ROS 2 node.

$58

this phase

  1. 3.1

    Workspaces, colcon and Overlays

    Build a workspace, source the overlay in a fresh terminal, and see what breaks when you do not.

  2. 3.2

    Your First Package

    ros2 pkg create, package.xml, rosdep, and an entry point that actually launches.

  3. 3.3

    Publisher and Subscriber in rclpy

    A node, a timer, a callback, and two programs you wrote exchanging messages.

  4. 3.4

    Your Own Interfaces and Parameters

    Custom msg and srv files, declared parameters, and a node you can retune from the CLI.

  5. 3.5

    Closing the Loop in a Node

    Move your Python controller into a node, launch it with parameters, and reach the goal.

By the end of this phase:A colcon package whose rclpy node reads a goal from a parameter and drives the robot to it.
4

A Body and a World

Describe a robot precisely enough that the simulator, RViz and a planner all agree where it is.

$58

this phase

  1. 4.1

    Frames and tf2

    base_link, odom and map, and reading the transform tree with tf2_echo when it breaks.

  2. 4.2

    URDF and Xacro

    Links, joints, inertia, and a robot model RViz renders without a single warning.

  3. 4.3

    Gazebo Harmonic

    SDF worlds, spawning your robot, and why Gazebo Classic stopped being an option in 2025.

  4. 4.4

    The ros_gz Bridge

    Get simulated lidar, IMU and odometry out of Gazebo and onto real ROS topics.

  5. 4.5

    Odometry and Its Drift

    Wheel odometry, an EKF from robot_localization, and measuring exactly how wrong you are.

By the end of this phase:Your own URDF robot driving in Gazebo Harmonic with a clean TF tree and live lidar in RViz 2.
5

Autonomy, Then Reality

Give the robot a goal instead of a velocity, then price what a physical one costs.

$58

this phase

  1. 5.1

    Mapping with slam_toolbox

    Build a map from lidar and odometry while driving, then save it to disk.

  2. 5.2

    Localizing in a Known Map

    AMCL, setting an initial pose, and telling a lost robot apart from a drifting one.

  3. 5.3

    Navigating with Nav2

    Costmaps, planner, controller, behavior tree, and a goal sent from RViz.

  4. 5.4

    When Nav2 Gets Stuck

    Recovery behaviors, costmap tuning, and diagnosing a failed run from its recorded bag.

  5. 5.5

    Off the Simulator

    micro-ROS on a microcontroller base, ros2_control, calibration, and an E-stop within reach.

  6. 5.6

    What Learning Adds, and What It Costs

    Imitation learning, RL and sim-to-real named honestly, with the GPU bill each one brings.

By the end of this phase:A robot that maps an unknown world and navigates it unattended, plus one recorded failure with a written cause.

Tools you will use

  • ROS 2 Jazzy
  • colcon
  • rclpy
  • Gazebo Harmonic
  • RViz 2
  • slam_toolbox
  • Nav2
  • micro-ROS

What you will build

  • The Robot You Described

    Your URDF in Gazebo Harmonic, lidar live in RViz 2

  • Map, Then Go

    slam_toolbox into Nav2, hands off the keyboard

  • One Real Robot

    A micro-ROS base taking cmd_vel and publishing real odometry