👉 Start with basic robotics introduction → and continue with the followings key modules.
- 1. Course Overview
- 2. Who This Course Is For
- 3. What You Will Learn
- Module 4 : Practical Robotics Math
- Module 5 : Robot Perception & Computer Vision
- Module 6 : Module 6: Mastering ROS 2 (Jazzy Jalisco): The Ultimate Industry Standard Middleware Guide
- 🛒 Buy Robotics Kits
- Join & start your AI & Robotics journey with UDHY.
1. Course Overview
This Advanced Robotics course is designed for learners who want to move beyond fundamentals and build real-world, industry-ready robotic systems.
The program covers robot perception, sensor fusion, kinematics, embedded systems, and AI-driven robotics, with a strong focus on hands-on projects and real applications such as autonomous vehicles, mobile robots, and intelligent machines.
2. Who This Course Is For
Computer Science / Engineering students
Robotics & AI learners moving beyond basics
Developers transitioning into Autonomous Vehicles (AV)
Engineers interested in ROS, perception, and robotics systems
3. What You Will Learn
Robot Kinematics (Forward & Inverse)
Sensor Fusion (LiDAR, Camera, IMU)
Robot Perception & Computer Vision
Embedded Systems & Real-Time Control
ROS (Robot Operating System)
Motion planning & control systems
👉 These are industry-standard skills used in robotics & AV [ Go to Top ]
Module 4 : Practical Robotics Math
Practical robotics math sounds scary, but this is about how a robot moves, senses, and makes decisions. You don’t need to be a math genius to start; you just need to understand a few core ideas. It is fundamental as it provides the language and tools robots need to understand, move, and interact with the world. Without math, concepts like kinematics, perception, and control systems cannot be modeled, simulated, or implemented effectively
This section covers the essential robotics math frameworks required to program and control robots. From how a robot calculates its position in 3D space to how it predicts its future path, we break down complex formulas into practical, buildable steps.
4.1. Applied Linear Algebra for Robotics: Mastering Rotation Matrices and Vectors
Linear Algebra is the language robots use to understand where they are in space. By applying matrices — grids of numbers — engineers can represent a robot’s position, orientation, and movement in a 3D world. Advanced learners will master vectors, matrices, and coordinate transformations, turning abstract math into industry‑ready skills. These concepts are the DNA of robotics: without them, a robot cannot know where its hand is or which way its sensors are pointing.
For example : If a robot arm is at $(x, y, z)$, how do we calculate the new coordinates when it rotates 30° to the left? We use a Rotation Matrix.
The Main Tools in Linear Algebra [ Go to Top ]
4.1.1. Vectors (Position and Direction)
A vector is just a list of numbers that tells the robot a distance and a direction. In robotics, a vector doesn’t just tell you “where” something is; it tells you “how much” and “in what direction.” We use vectors to represent a robot’s velocity, the force a finger applies to an object, or the distance to an obstacle detected by a sensor.
Simple Info: It represents a point in space (X, Y, Z).
Example: If your drone is flying at 5 meters per second toward the North-East, that is a 2D vector. In 3D robotics, we use $x, y,$ and $z$ coordinates to pinpoint any spot in the room relative to the robot’s “home” position. Or to tell a drone to fly 5 meters forward and 2 meters up, you use a vector like [5, 0, 2].
Why it matters: Every time a robot moves, it is calculating “Vector Addition” to combine its current position with its intended movement.
4.1.2. Matrices (Changing the View)
A matrix is a grid of numbers (like a small spreadsheet). In robotics, we use them to change vector. A matrix is a rectangular grid of numbers. In robotics, we don’t just use them to store data; we use them as “Operators.” A matrix can act like a set of instructions that tells a vector to grow, shrink, or rotate.
Simple Info: It can rotate, stretch, or move a vector from one place to another.
Example: Imagine a 3×3 matrix. This grid can store the “Rotation” of a robot’s camera. When the camera data (a vector) is multiplied by this matrix, the image “rotates” in the robot’s brain so it stays level even if the robot is tilting. or If a robot arm rotates 90 degrees, a “Rotation Matrix” calculates the new position of every part of that arm instantly.
Why it matters: Modern robotics (and AI) runs almost entirely on Matrix Multiplication. It is the fastest way for a computer to process thousands of spatial calculations at once. [ Go to Top ]
4.1.3. Transformations (Moving Parts)
This combines rotation and movement into one step. This is the most critical concept in robotics. It is the math of “Perspective.” A robot has many “Coordinate Frames”: the World frame (the room), the Base frame (the robot’s feet), and the Tool frame (the robot’s hand).
Simple Info: It maps how one part of the robot relates to another.
Example: A “Transformation Matrix” tells the robot: “The camera is 10cm above the wheels. If the wheels move forward, here is where the camera is now.”Your camera sees a “Cup” at coordinates $(1, 2, 5)$ relative to the camera. But the robot’s hand is 2 meters away from the camera. Coordinate Transformation is the math that translates the cup’s position so the hand knows exactly where to grab.
Why it matters: Without transformations, a robot could see an object but would never be able to touch it because it wouldn’t know where the object is relative to its own arm. [ Go to Top ]
4.2. Kinematics (forward & inverse)
Kinematics is often called “the geometry of motion” because it describes how a robot moves through space without worrying about the forces (like weight or friction) that cause the movement. In robotics engineering, we split this into two main problems: Forward Kinematics (FK) and Inverse Kinematics (IK) .
4.2.1. Forward Kinematics (FK): “Where is my hand?”
The Concept: You know the angles of every motor (joint) in the robot arm, and you want to calculate exactly where the tip of the robot (the “End-Effector”) is in 3D space $(x, y, z)$.
The Math: We use the Denavit-Hartenberg (D-H) Parameters. This is a standard way of labeling robot joints and links so we can use a single formula to find the position.
Real-World Example: If a robot knows joint 1 is at 45° and joint 2 is at 90°, Forward Kinematics tells the computer: “Your gripper is currently 1.2 meters high and 0.5 meters forward.” [ Go to Top ]
4.2.2. Inverse Kinematics (IK): “How do I get there?”
The Concept: This is much harder. You know where you want the robot to go (e.g., “Pick up that apple at coordinates $1, 2, 5$”), and you need to calculate what angles the motors need to turn to get there. Advanced learners must also understand Singularities—specific configurations where the robot’s math ‘breaks’ and joints become locked or uncontrollable.
The Challenge: Sometimes there is more than one way to reach a point. Imagine touching your nose—you can do it with your elbow down or your elbow up. The robot has to choose the “best” way.
Real-World Example: This is how a surgical robot works. The surgeon moves a joystick to a position $(x, y, z)$, and the Inverse Kinematics instantly calculates the motor movements to put the scalpel in that exact spot. [ Go to Top ]
4.3. Robotics Motion calculations
Motion calculations are the math used to plan how a robot gets from Point A to Point B. While kinematics (which we discussed) focuses on positions, motion calculations focus on time, speed, and smoothness. It is the difference between a robot “snapping” into a position and a robot moving gracefully like a human.
The 3 Big Ideas in Motion
Velocity (Speed): How fast is the robot moving right now?
Acceleration: How quickly is the robot speeding up or slowing down?
Jerk: How “shaky” is the movement? Lowering “jerk” makes the robot last longer because it doesn’t vibrate the parts.
Hence, Motion calculations is the “nervous system” of a robot. It is the process of breaking down a high-level task (like “pick up that cup”) into thousands of tiny mathematical steps. For a robot to move successfully, it doesn’t just need a destination; it needs a safe path and a smooth plan. In motion calculations in robotics do include concepts like Path Planning, Trajectory Generation, PID Control (The Correction) and Odometry (Wheel Counting). These are advanced layers built on top of kinematics and dynamics, and they are critical for autonomous robots to operate safely and intelligently. [ Go to Top ]
4.3.1. Path Planning (The Map)
Path Planning: Focuses on finding a collision‑free route from start to goal. This is the “static” map. It’s a list of $(x, y, z)$ points the robot should follow to avoid walls. It doesn’t care about time or speed.
This is finding the “line” the robot should follow to avoid hitting walls or people.
Simple Info: It’s like a GPS for the robot. It finds the shortest or safest path.
Example: A warehouse robot sees a box in the hallway and calculates a curved path to drive around it. or A GPS map showing the blue line you should drive on. [ Go to Top ]
4.3.2. Trajectory Generation (The Timing)
Trajectory Generation: Adds timing and dynamics, defining how fast and smoothly the robot should move along the path. This is the “dynamic” plan. It adds time, velocity, and acceleration to the path. Once the robot has a path (the line), it needs a schedule. It decides when to speed up and when to brake.
Simple Info: It adds time to the path.
Example: A robot arm shouldn’t start at full speed. It calculates a “S-curve” to start slow, move fast in the middle, and slow down before it stops. or Deciding how fast to drive on that blue line so you don’t skid on a turn or break the engine. [ Go to Top ]
4.3.3. PID Control (The Correction)
PID Control is a math formula that helps a robot reach a target exactly without overshooting it or shaking. it focuses on Accuracy and Smoothness and constantly “checks” the error (the gap between where the robot is and where it wants to be) and fixes it.
P (Proportional): The “Big Push.” If the robot is far away, move fast. If it’s close, move slow.
I (Integral): The “Stubborn Fix.” If the robot is stuck (like on a carpet), this adds power over time to get it moving.
D (Derivative): The “Brakes.” This looks at how fast the robot is closing the gap and slows it down so it doesn’t crash into the target.
This math constantly checks for mistakes while moving.
Simple Info: It compares where the robot should be with where it actually is and fixes it.
Example 1:
A drone feels a gust of wind. The PID math detects it is tilting and instantly spins the motors faster to stay level.
Example 2:
– Imagine a self-balancing robot (like a Segway).
– If it tips forward, the PID math detects the angle.
– It tells the wheels to move forward to catch the fall.
– Without PID, the robot would over-correct, wobble back and forth, and eventually fall over. With PID, it stays perfectly still. [ Go to Top ]
4.3.3. Odometry (Wheel Counting)
Odometry is how a robot “estimates” its position by focusing position tracking/counting how many times its wheels have spun. It uses sensors called Encoders to count the clicks of the motor.
The Math: If you know the circumference (distance around) of the wheel, you can multiply:
Distance = (Number of Spins) × (Wheel Circumference)
This is calculating motion based on how much the wheels have turned.
Simple Info: If the robot knows the wheel size, it counts spins to guess how far it has moved.
Example 1:
A robot car counts 10 wheel rotations. Based on the wheel size, it calculates it has moved exactly 2 meters forward.
Example 2:
Imagine a vacuum robot (like a Roomba) in a dark room where it can’t see.
– The robot knows its wheels are 20cm around.
– The encoders count 10 full spins.
– The robot calculates: “I have moved exactly 2 meters forward.”
– The Problem: If the wheels slip on a rug, the robot thinks it moved 2 meters, but it might have only moved 1 meter. This is why odometry is usually combined with other sensors. [ Go to Top ]
Module 5 : Robot Perception & Computer Vision
Robot perception is important because it gives machines the ability to sense and interpret the world around them. Without perception, a robot is essentially blind — it cannot recognize objects, understand its environment, or make safe decisions. By combining inputs from cameras, LiDAR, radar, and IMUs, robots build a reliable model of their surroundings, enabling tasks like navigation, manipulation, and human interaction.
👉 This topic has been discussed in detail on UDHY.com other courses, where you’ll find structured modules to help learners turn theory into industry‑ready skills.
This section covers the essential Robot Perception, that gives Machines the Power of Sight and Understanding. This section covers how a robot processes raw data from cameras and sensors to recognize objects, map its surroundings, and make intelligent decisions in real-time. [ Go to Top ]
5.1. Image Processing & Traditional Vision
Image processing and traditional computer vision are the foundations of robotic perception. Before deep learning became dominant, robots relied on classical vision techniques to interpret their environment. These methods remain essential today because they are lightweight, fast, and effective for many real‑world tasks.
In Simple English: It’s like giving the robot a pair of glasses. We use math to make the blurry images clear and tell the robot, “This straight line is the edge of a table.” Image processing and traditional vision are the DNA of robotic perception. They provide the mathematical and algorithmic tools robots need to interpret their environment, enabling navigation, manipulation, and interaction.
Example: A vacuum robot identifying the line where the floor meets the wall so it doesn’t crash.
Key Skills: Filtering, Edge Detection (Canny), and Color Segmentation. [ Go to Top ]
5.2. Object Detection & Recognition
Object detection and recognition are the core AI skills that give robots true intelligence. Unlike traditional vision methods, AI‑powered detection allows robots to identify, classify, and track objects in real time — enabling safe navigation, manipulation, and human interaction.
In Simple English: We show the robot millions of pictures of “cats” and “dogs.” Eventually, the robot learns to spot them on its own, even in the dark or from a weird angle. Object detection and recognition are the AI backbone of robotics perception. They transform raw sensor data into actionable intelligence, enabling robots to see, understand, and interact with the world.
Example: A security robot recognizing a “human” versus a “moving box” in a warehouse.
Key Skills: YOLO (You Only Look Once), CNNs (Convolutional Neural Networks), and TensorFlow/PyTorch. [ Go to Top ]
5.3. 3D Vision & Depth Sensing
Vision and depth sensing are the eyes of a robot, enabling it to perceive its environment in 2D and 3D. While cameras provide color and texture information, depth sensors like LiDAR, stereo vision, and structured light allow robots to measure distance, detect obstacles, and understand spatial geometry. Together, they form the foundation of safe navigation, manipulation, and human‑robot interaction.
In Simple English: This gives the robot “depth perception,” just like humans. It allows the robot to know exactly how many centimeters away an object is. Vision and depth sensing are the core perception skills every robotics engineer must master. They allow robots to see, measure, and interact with the world — transforming raw sensor data into actionable intelligence for navigation, manipulation, and safety.
Example: A drone flying through a forest needs 3D vision to weave between branches without hitting them.
Key Skills: Stereo Matching, Point Clouds, and LiDAR Data Processing. [ Go to Top ]
5.4. Semantic Segmentation
Semantic segmentation is the process of classifying every pixel in an image into meaningful categories. Unlike object detection, which draws bounding boxes, segmentation gives robots a pixel‑level understanding of their environment. This is critical for tasks like autonomous driving, robotic manipulation, and human‑robot interaction, where precision and context matter.
In Simple English: Instead of just seeing a “blob,” the robot colors its entire world view like a map: “These pixels are Road,” “These pixels are Sidewalk,” and “These pixels are Sky.” Semantic segmentation is the AI backbone of robotic perception, enabling robots to interpret environments with pixel‑level precision. It transforms raw images into actionable intelligence, powering navigation, manipulation, and safety in advanced robotics systems.
Example: A self-driving car needs to know exactly where the drivable road ends and the sidewalk begins.
Key Skills: Image Masking and Scene Understanding. [ Go to Top ]
Module 6 : Module 6: Mastering ROS 2 (Jazzy Jalisco): The Ultimate Industry Standard Middleware Guide
The Robot Operating System (ROS) is the essential backbone of modern robotics. As a powerful open-source middleware framework, ROS standardizes how complex machines communicate, process sensor data, and execute tasks. By utilizing global libraries for perception and control, engineers avoid “reinventing the wheel,” allowing for faster development and seamless compatibility across diverse hardware platforms.
Step Into Professional Development with ROS 2 Jazzy Jalisco
In 2026, ROS 2 is the undisputed “Android of Robotics.” This course skips the outdated, centralized “Master” node of ROS 1 to focus on the decentralized, high-performance architecture of Jazzy Jalisco.
Key Skills You Will Master:
– Modern Architecture: Dive deep into Nodes, Topics, and Lifecycle Management for robust system stability.
– Autonomous Navigation: Implement Nav2 (Navigation 2) for sophisticated path planning and obstacle avoidance.
– High-Precision Control: Leverage MoveIt 2 for expert-level robotic arm manipulation and motion planning.
– Reliable Communication: Understand how DDS (Data Distribution Service) ensures your robot stays connected in real-world industrial environments.
Whether you are building autonomous vehicles or smart factory arms, mastering ROS 2 is the single most critical step toward becoming a career-ready robotics developer. [ Go to Top ]
6.1. The ROS 2 Architecture: A Distributed Powerhouse
Table below shows What is the difference between ROS 1 and ROS 2?
| Feature | ROS 1 (Legacy) | ROS 2 (Modern Standard) |
| Communication Layer | Custom TCPROS/UDPROS (Centralized) | DDS (Data Distribution Service) (Decentralized) |
| Master Node | Required (ROS Master) | No Master (Nodes discover each other) |
| Real-Time Support | No (Best effort only) | Native Support for real-time systems |
| Platform Support | Mainly Linux (Ubuntu) | Linux, Windows, and macOS |
| Connectivity | Requires stable, wired networks | Built for noisy wireless/Wi-Fi networks |
| Security | None (Open by default) | Secure ROS (SROS) with encryption/auth |
| API Architecture | Monolithic | Modular (Internal API and Client API) |
ROS 1 vs. ROS 2: Architectural Comparison
The primary difference between the two versions is the transition from a centralized system (one manager) to a decentralized system (distributed intelligence). This shift allows ROS 2 to be used in high-stakes environments like autonomous driving and multi-robot swarms where a single point of failure is not acceptable.
Unlike the centralized ROS 1, ROS 2 architecture is built for the future: multi-robot swarms, real-time precision, and reliable performance over wireless networks. It has removed the “Single Point of Failure” to become the industrial standard for 2026.
The Core: Data Distribution Service (DDS) : ROS 2 replaces the old “Master” node with DDS, a decentralized discovery layer.
Why it matters: Nodes now find each other automatically. If one part of the system fails, the rest stays online.
QoS (Quality of Service): You can now prioritize critical data (like Emergency Stops) to ensure they arrive first, even on laggy networks.
[ Go to Top ]
6.1.1. ROS 2 Key Components for Advanced Developers
To build career-ready systems, you must master these four pillars:
1. Nodes & Lifecycle Management Individual processes (Nodes) now have Lifecycle states (Unconfigured, Inactive, Active). This allows you to manage power and safety by starting or stopping sensors without crashing the robot.
2. Topics: High-Frequency Data The anonymous Publish/Subscribe model allows for a modular design. Swap a 2D LiDAR for a 3D one without rewriting your entire code base.
3. Services & Actions: Logic & Goals
– Services: Quick “Request/Response” calls (e.g., Check Battery).
– Actions: Complex, long-running goals (e.g., Navigate to Room A) that provide constant feedback and can be canceled mid-move.
4. Distributed Parameters Settings are now stored locally on each node, preventing naming conflicts and making your system infinitely more scalable.[ Go to Top ]
6.2. ROS 2: How to Get Started (2026 Edition)
Starting with ROS 2 (Robot Operating System) can be challenging due to strict environment requirements. To build career-ready skills, follow this optimized roadmap to set up a professional development workspace.
Step 1: The Environment (Linux is Industry Standard)
For 2026 robotics development, Ubuntu Linux 24.04 LTS is the essential foundation.
Pro Tip: If you are on Windows, avoid a direct “bare metal” installation. Instead, use WSL 2 (Windows Subsystem for Linux) or Docker containers. This allows you to run a full Linux kernel for ROS 2 without partitioning your hard drive, ensuring a clean and error-free environment.
Step 2: Installing ROS 2 Jazzy Jalisco
The current stable “Long Term Support” (LTS) version is ROS 2 Jazzy Jalisco.
Installation: Open your terminal and run: sudo apt install ros-jazzy-desktop.
Environment Sourcing: ROS 2 functions through specific environment variables. You must “source” your workspace in every new terminal so the system can locate ROS commands: source /opt/ros/jazzy/setup.bash (Tip: Add this line to your .bashrc file to automate this step!)
Step 3: Your First Project (Turtlesim)
Turtlesim is the industry-standard “Hello World” for robotics. It teaches you how the Computational Graph functions.
Start the Simulator: ros2 run turtlesim turtlesim_node
Launch the Controller: In a separate terminal, run: ros2 run turtlesim turtle_teleop_key.
Analyze the Logic: While driving the turtle, run ros2 topic list. This reveals the “Topics” (the data wires) connecting your keyboard input to the robot’s motors.
The “Professional Toolbox” for Success
To move beyond basics, you must master the three tools that define a professional robotics workflow:
– The Build Tool (Colcon): Think of Colcon as your project manager. It compiles your C++ and Python code into executable packages, managing all dependencies automatically.
– The Visualizer (RViz2): This is your 3D Debugger. RViz2 allows you to “see” through the robot’s eyes—visualizing LiDAR point clouds, camera feeds, and coordinate frames (TF) in a 3D space.
– The Simulator (Gazebo Harmonic): For high-stakes testing, move from Turtlesim to Gazebo. This physics engine simulates real-world gravity, friction, and sensor noise, allowing you to test autonomous algorithms without risking expensive hardware. [ Go to Top ]
Getting Started with Robotics
To practice 3D sensing, we recommend our AI-compatible Depth Camera kit 
.
🛒 Buy Robotics Kits
Ready to start building your first robot? Visit UDHY’s Robotics Online Store to explore various robotics kits designed for learning sensors, motors, and coding. Each kit includes everything you need to build, test, and understand real robots—perfect for students, hobbyists, and future innovators. [ Go to Top ]
Join & start your AI & Robotics journey with UDHY.
Enter your email address to register to our newsletter subscription delivered on regular basis!
