Naar de inhoud

De hoofdinhoud van deze pagina is in het Engels.Deze pagina vertalen

AI & trending onderwerpen

Robotics roadmap

Robotics combines electronics, programming, maths and now AI. Robots that understand language and vision are one of the hottest areas in 2026. This roadmap goes from microcontrollers to ROS 2, simulation and AI-powered robots.

Tijd
⏱ 6–9 months
Level
Intermediate
Steps
6 steps
Architecture ↓Cheat sheet ↓4 interview Q&As ↓
Delen

The roadmap

Click a step to open the full guide.

Best architecture: a ROS 2 mobile robot

The layered software stack of an autonomous robot.

Robot software stack

  1. Mission / task planner (optionally LLM)
  2. Navigation (SLAM, path planning)
  3. Perception (camera, lidar, detection)
  4. Control (PID, motor drivers)
  5. Hardware + independent safety (e-stop)

Why it's built this way

  • Each layer is a set of ROS 2 nodes talking over topics.
  • Real-time control runs close to hardware, separate from AI.
  • Safety is independent of all software above it.

Robotics cheat sheet

The essentials on one screen. Bookmark it for revision.

Concepts

Ohm's lawV = I × R
PWMControl motor speed/servo angle
PIDP = present error, I = past, D = future trend
Forward kinematicsJoint angles → end position
Inverse kinematicsTarget position → joint angles
SLAMMap + localise at the same time

ROS 2 commands

List topicsros2 topic list
See messagesros2 topic echo /scan
Run a noderos2 run pkg node
Launchros2 launch pkg file.launch.py
Build workspacecolcon build

Improve performance

  • Keep real-time control loops on microcontrollers or real-time threads, not in slow AI code.
  • Run AI models on edge accelerators and quantise them.
  • Reduce sensor data rates and resolution to what you need.
  • Profile ROS 2 message rates and latency; avoid huge messages.

Security checklist

  • Safety first: emergency stop, speed and force limits independent of software.
  • Secure robot networks; don't expose ROS ports to the internet.
  • Authenticate remote control and updates; sign firmware.
  • Log commands for incident review.

Most important interview questions

Click a question to see a strong answer. Each step also has its own questions.

1What are the main parts of a robotic system?

Sensing (sensors), perception (understanding data), planning (deciding), control (executing motions) and actuation (motors), plus power and safety systems.

2Why is ROS 2 used?

It provides standard communication, tools and libraries (navigation, visualisation, drivers) so teams can build complex robots from reusable parts.

3How do you tune a PID controller?

Start with P until it oscillates slightly, add D to damp overshoot, then add a little I to remove steady-state error; test across conditions.

4Wat is sensor fusion?

Combining data from multiple sensors (e.g. IMU, wheel encoders, GPS) to get a more accurate and robust estimate than any one sensor.

Related roadmaps

6

IoT (Internet of Things)

6 steps · 4–6 months · Beginner

Connect devices, collect data and automate the physical world.

View roadmap →
6

AI Basics & the AI Family Tree

6 steps · 2–3 weeks · Beginner

How AI, ML, deep learning, generative AI and agents fit together.

View roadmap →
6

Python

6 steps · 6–8 weeks · Beginner

The language of AI, data and automation.

View roadmap →
1

Robotics · Step 1 of 6 · 4 weeks

Electronics and microcontrollers

Sensors, motors and the code that drives them.

Every robot needs to sense the world and move in it. Microcontrollers such as Arduino and ESP32 read sensors and control motors in real time.

Start with a simple wheeled robot kit; it teaches wiring, power and control without much cost.

What you'll learn

  • Voltage, current and circuits
  • Arduino and ESP32
  • Sensors (distance, IMU, encoders)
  • DC, servo and stepper motors
  • Motor drivers
  • Power and batteries

Sense → think → act

  1. Sensors
  2. Microcontroller
  3. Decision
  4. Motor driver
  5. Motors

Key concepts

Ohm's law
==V = I × R==: voltage = current × resistance.
PWM
Fast on/off switching to control speed or position.
Motor driver
A chip that lets a small microcontroller safely power motors.

Step by step

  1. 1.1

    Circuits

    Voltage, current, resistance (Ohm's law), breadboards.

  2. 1.2

    Microcontroller

    Blink an LED, read a button, use the serial monitor.

  3. 1.3

    Sensors

    Ultrasonic distance, IMU, wheel encoders.

  4. 1.4

    Motors

    Drive DC motors through a motor driver; control servos with PWM.

Notes

  • Most beginner hardware problems are power or wiring issues.

Tips & tricks

  • Buy a multimeter early; it saves hours of guessing.

Interview questions

Why can't you power a motor directly from a microcontroller pin?

Pins supply very little current; motors need much more and create electrical noise and voltage spikes. Use a motor driver and a separate power supply.

The quick way to study this

Build a line-following or obstacle-avoiding robot from a kit.

IoT roadmap →
2

Robotics · Step 2 of 6 · 4 weeks

Maths and control

Move precisely.

Control theory makes robots move smoothly. PID controllers are used everywhere, from wheels to drones. Kinematics explains how joint angles relate to where a robot arm's hand ends up.

Sensor data is noisy; filters such as the Kalman filter combine readings into better estimates.

What you'll learn

  • Coordinate frames and transforms
  • Kinematics
  • PID control
  • Basic linear algebra
  • Sensor noise and filtering

PID control loop

  1. Target value
  2. Measure actual
  3. Error = target − actual
  4. PID output
  5. Apply to motor

Repeats as a loop

Key concepts

PID controller
Corrects error using now (P), past (I) and trend (D).
Inverse kinematics
Working out joint angles to reach a target point.
Kalman filter
Combines noisy measurements into a better estimate.

Step by step

  1. 2.1

    Frames

    World, robot and sensor coordinate frames.

  2. 2.2

    Kinematics

    Forward (angles → position) and inverse (position → angles).

  3. 2.3

    PID

    Proportional, integral, derivative control.

  4. 2.4

    Filtering

    Smooth noisy sensors; combine them (sensor fusion).

Notes

  • Tune PID one term at a time: P, then D, then I.

Tips & tricks

  • Plot sensor readings live to understand noise.

Interview questions

Explain PID control simply.

P pushes harder the further you are from the target, I fixes small steady errors over time, D slows down to avoid overshooting. Together they reach the target smoothly.

The quick way to study this

Implement a PID speed controller on your robot.

3

Robotics · Step 3 of 6 · 5 weeks

ROS 2 and Linux

The standard robotics software framework.

ROS 2 (Robot Operating System) is a framework of tools and libraries used in research and industry. Programs (nodes) communicate through topics and services, so you can mix sensors, planners and controllers.

It runs best on Ubuntu Linux.

What you'll learn

  • Linux basics
  • ROS 2 nodes, topics, services and actions
  • Launch files
  • TF2 transforms
  • Debugging tools (RViz)

ROS 2 nodes talking over topics

/cmd_vel, /scan, /odom topics
  • Lidar driver node
  • Navigation node
  • Motor controller node
  • Camera node

Key concepts

Node
One program in a ROS system.
Topic
A named channel for streaming messages (publish/subscribe).
Service vs action
Service: quick request/reply. Action: long task with feedback.

Step by step

  1. 3.1

    Ubuntu + ROS 2

    Install a supported ROS 2 distribution.

  2. 3.2

    Nodes and topics

    Publisher/subscriber programs in Python.

  3. 3.3

    Services and actions

    Request/response and long-running goals.

  4. 3.4

    Tools

    RViz for visualisation, ros2 topic echo voor debugging.

Notes

  • Learn ROS 2, not the original ROS 1, which is end-of-life.

Tips & tricks

  • Follow the official beginner tutorials in order.

Interview questions

Topic vs service in ROS 2?

Topics are continuous, many-to-many streams (sensor data). Services are one-off request/response calls (e.g. 'reset odometry').

The quick way to study this

Complete the ROS 2 beginner tutorials and write two nodes that talk to each other.

4

Robotics · Step 4 of 6 · 3 weeks

Simulation

Test robots safely and cheaply.

Simulators let you test code without breaking hardware. Describe your robot in a model file, add simulated sensors and run your ROS 2 code against it.

Expect differences between simulation and reality, and test on real hardware regularly.

What you'll learn

  • Gazebo or similar simulators
  • Robot models (URDF)
  • Simulated sensors
  • Sim-to-real gap

Simulation vs real robot

Simulation

  • Free, safe, fast
  • Perfect repeatability
  • Physics approximated

Real robot

  • Costs and can break
  • Noise, wear, latency
  • The real test

Key concepts

URDF
A file describing the robot's parts and joints.
Sim-to-real gap
Differences between simulation and reality.

Step by step

  1. 4.1

    Robot model

    Describe links and joints in URDF.

  2. 4.2

    Simulate sensors

    Add a lidar and camera in the simulator.

  3. 4.3

    Test code

    Run the same ROS 2 nodes against the simulated robot.

Notes

  • Simulation is great for logic; real hardware reveals physics problems.

Tips & tricks

  • Use a simulated version of a popular robot to learn.

Interview questions

How do you reduce the sim-to-real gap?

Model sensor noise and delays, randomise physics parameters in simulation, calibrate the real robot, and test on hardware early and often.

The quick way to study this

Drive a simulated robot around a map with ROS 2.

5

Robotics · Step 5 of 6 · 5 weeks

Perception and navigation

Let robots see and find their way.

Perception turns sensor data into understanding: where obstacles are, what objects are present. SLAM builds a map while tracking the robot's position. Planners then find safe paths.

ROS 2's navigation stack provides many of these pieces ready to configure.

What you'll learn

  • Cameras and LiDAR
  • OpenCV basics
  • Object detection
  • SLAM (mapping)
  • Path planning
  • Navigation stack

Autonomous navigation

  1. Sensors
  2. Perception
  3. Localisation (SLAM)
  4. Path planning
  5. Control
  6. Motors

Key concepts

SLAM
Simultaneous localisation and mapping: make a map and know where you are in it.
Costmap
A grid showing where the robot can and can't safely go.

Step by step

  1. 5.1

    Perception

    Cameras, depth and lidar; object detection.

  2. 5.2

    SLAM

    Build a map while localising.

  3. 5.3

    Planning

    Global path + local obstacle avoidance.

Notes

  • Good calibration of cameras and sensors matters more than clever algorithms.

Tips & tricks

  • Start with a pre-trained object detection model.

Interview questions

Wat is SLAM and why is it hard?

Building a map while estimating the robot's position in it. It's hard because both are uncertain and errors compound; loop closure corrects drift.

The quick way to study this

Map a room in simulation and navigate to a goal point.

6

Robotics · Step 6 of 6 · 4 weeks

AI for robots

Language, vision and learning.

Modern robots increasingly use AI models that understand images and instructions, and learn skills from demonstrations. LLMs can turn spoken requests into step-by-step plans that the robot's planner executes.

Safety is critical: keep hard limits and emergency stops outside the AI model.

What you'll learn

  • Vision-language models for robots
  • Imitation learning
  • Reinforcement learning basics
  • LLMs for task planning
  • Edge AI hardware
  • Safety

AI robot stack

  1. Language / task planner (LLM)
  2. Perception (vision models)
  3. Motion planning
  4. Low-level control (PID)
  5. Safety layer (hard limits, e-stop)

Key concepts

Imitation learning
Learning a skill by copying demonstrations.
Reinforcement learning
Learning by trial and reward, often in simulation first.
Edge AI
Running models on the robot itself.

Step by step

  1. 6.1

    Vision-language models

    Robots that understand images and instructions.

  2. 6.2

    Learning from demonstration

    Imitation learning from human examples.

  3. 6.3

    LLM planning

    Turn a spoken request into steps the planner executes.

  4. 6.4

    Safety

    Hard limits and emergency stops outside the AI.

Notes

  • Never let an AI model directly bypass physical safety limits.

Tips & tricks

  • Run small models on edge devices such as Jetson or a Raspberry Pi with an accelerator.

Interview questions

How would you use an LLM safely in a robot?

Use it for high-level planning only, convert its output into a validated set of allowed actions, keep safety limits and emergency stops in independent low-level code, and require confirmation for risky tasks.

The quick way to study this

Make your robot follow simple spoken commands using an LLM planner.

AI basics →

Deel deze pagina