Robotics · Free course
Robotics and AI
Learn the maths, control and machine learning that let a robot arm see, plan and learn tasks from demonstrations, then deploy it safely.
Robots are moving from fixed, hand-programmed motions to behaviours learned from data. A low-cost arm, a camera and a few dozen recorded demonstrations are now enough to train a policy that picks and places objects. Doing that well still depends on the classical foundations: coordinate frames, kinematics, control and camera geometry.
This course covers both halves. It starts with the maths of rotations and transforms, builds forward and inverse kinematics and joint control for a servo-driven arm, adds camera geometry and deep-learning perception, and then moves to imitation learning, reinforcement learning, sim-to-real transfer and vision-language-action models. It ends with deployment on edge hardware, where an FPGA can help, and the safety checks a learned controller needs.
Who it is for
Engineering and computer science students, software engineers moving into robotics, and engineers who want a structured path from classical robotics to robot learning.
Hardware
No hardware is needed for the maths, control and learning examples; every one runs in Python on a laptop. The hands-on labs use a pair of SO-101 arms (a leader arm for teleoperation and a follower arm), one or two USB cameras and, for training, a GPU workstation or cloud GPU.
Software
Python 3 with NumPy, SciPy and Matplotlib for the worked examples, and Hugging Face LeRobot for teleoperation, data recording and policy training on the arm.
Syllabus
5 modules · 12 lessons · 10 h 35 minFoundations
2 lessons · 1 h 25 min- 1Robots, AI and the SO-101 armWhat a robot is, how learned behaviour differs from programmed motion, and how the low-cost SO-101 arm, its STS3215 servos and LeRobot fit together safely.beginner35 min
- 2Spatial maths for robots: rotations and transformsRotation matrices, roll-pitch-yaw and gimbal lock, axis-angle, quaternions and SLERP, homogeneous transforms, and how a neural network should output a rotation.intermediate50 min
Kinematics and control
3 lessons · 2 h 40 min- 3Forward kinematics for robot arms: DH, URDF and the SO-101Compute where a robot arm's gripper is from its joint angles: a planar 2-link arm by hand, DH tables, product of exponentials, and the SO-101's URDF in NumPy.intermediate55 min
- 4Inverse kinematics and Jacobians for robot armsSolve inverse kinematics for a 2-link arm in closed form, build the Jacobian, find singularities, and write damped least-squares IK and resolved-rate control.intermediate50 min
- 5Actuators, PID joint control and trajectoriesHow DC motors, gearboxes and smart bus servos drive a robot joint, how to write a discrete PID controller with anti-windup, and how to plan smooth trajectories.intermediate55 min
Perception
2 lessons · 1 h 45 min- 6Camera geometry for robots: pinhole model and calibrationProject 3D points with the pinhole model, correct lens distortion, calibrate a camera, and turn a pixel into a table position for a robot arm with homographies.intermediate50 min
- 7Deep learning for robot perception: CNNs and detectionHow CNNs find objects for a robot arm: convolution and pooling, detector outputs, IoU and NMS, precision, recall and mAP, and a 3D target from a 2D box.intermediate55 min
Robot learning
3 lessons · 2 h 55 min- 8Imitation learning: behaviour cloning and DAggerTrain a robot policy from demonstrations: behaviour cloning, compounding errors, DAgger, multimodal data, and recording an SO-101 dataset with LeRobot.intermediate1 h
- 9Action chunking and diffusion policies: ACT and beyondWhy robot policies predict action chunks, how ACT and diffusion policies work, training them with LeRobot, and reporting success rates with a Wilson interval.advanced1 h
- 10Reinforcement learning for robots: Q-learning to PPOMDPs and the Bellman equation, tabular Q-learning in NumPy, PPO's clipped objective, reward design for reaching, and why most robot RL runs in simulation.intermediate55 min
Deployment
2 lessons · 1 h 50 min- 11Sim-to-real transfer and vision-language-action modelsClose the reality gap with system identification and domain randomisation, then see how VLA models such as SmolVLA map images and instructions to robot actions.advanced55 min
- 12Deployment, edge inference, FPGAs and robot safetyBudget control-loop latency, time a servo bus, quantise a model to INT8, pick an edge target, and add software and hardware safety limits to a learned policy.intermediate55 min
Frequently asked questions
Do I need a robot arm to take this course?
No. The maths, control, vision and learning examples all run in Python on a laptop. An arm is only needed for the hardware labs, and the course explains what each lab would show if you do not have one.
What background do I need?
Python programming, basic linear algebra (vectors and matrices) and first-year calculus. Some exposure to machine learning helps for the second half but is not required.
Does the course use ROS 2?
The arm labs use the LeRobot Python tools, which keep the software stack small. ROS 2 is mentioned where it is the usual industry choice, and the concepts transfer directly.
Is the course free?
Yes. All lessons, quizzes and code downloads are free to use.
Latest Robotics and AI news
All news →- Video Friday: Drone Squeezes Itself to Fit Through Tight Spaces
- SECO’s AMD Ryzen™ Embedded 8000 Series-based COM Express module now available for sampling
- Robot Talk Episode 171 – Programming robot behaviour, with Sebastian Castro
- TI’s multiprotocol comms SoM connects factories
- Shaping the Form, Function, and Future of Humanoid Robotics
- SECO announces Modular Link IQ-615 and Modular Link MX93 are now available to order
Upcoming events
Calendar →Follow the Robotics and AI course
Get an email when new lessons, projects and exercises are added.