🤖 Robotics & Mechatronics

What you’ll learn

1. A Robot Is a Machine That Closes a LoopDefine a robot by its feedback loop and distinguish open-loop from closed-loop machines.A robot isn't defined by power or complexity but by continuously asking 'where am I versus where I should be?' — the error. Sensors measure it, actuators reduce it, control code decides how hard to push. Open-loop machines command and hope, which is fine when the world is predictable; closed-loop machines cost more and can go unstable, and what you buy is the ability to be wrong and recover.
2. Actuators: Turning Current Into TorqueExplain how motors make torque, why back-EMF shapes their behaviour, and compare brushed, brushless, stepper, and servo actuators.Torque is proportional to current — which is why a stalled motor makes maximum torque, zero power, and enormous heat. Back-EMF rises with speed and opposes the supply, producing the characteristic high-torque-at-low-speed curve. Steppers trade the sensor away for pulse counting, which is open-loop in disguise: exceed their torque and they skip steps silently while the controller keeps believing a false position.
3. Gearing: The Trade Nobody EscapesExplain the torque–speed trade in gearing, reflected inertia, and why backlash destabilises a control loop.A gearbox can't create power, so multiplying torque divides speed by the same factor (minus efficiency losses) — while multiplying reflected inertia by the ratio SQUARED. Backlash is the dead zone where the motor turns and the load doesn't, so the loop reads a lie and hunts. Strain wave (Harmonic Drive) gearing earns its cost by delivering high coaxial ratios with essentially zero backlash.
4. Encoders: How a Robot Knows ItselfCompare incremental and absolute encoders and explain why proprioception alone always drifts.Incremental encoders count pulses and use quadrature for direction, but measure change rather than position — so they need homing after power-up. Absolute encoders report a unique code per angle and use Gray code so misalignment can only ever cause an off-by-one error. Crucially, an encoder measures the motor, not the world: backlash, flex, and windup sit between them, and dead reckoning integrates errors that never cancel.
5. Sensing the WorldExplain why every sensing technology has a principled blindness, and why fusion is the answer.Ultrasonics are fooled by soft and angled surfaces; infrared by dark carpet; cameras lack depth and depend on light; lidar is defeated by glass, mirrors, and fog. You don't choose a sensor, you choose which failures you can live with. Fusion combines sensors whose failure modes don't overlap — and pairing a drifting-but-smooth estimate with a noisy-but-drift-free one (classically via a Kalman filter) beats either alone.
6. Feedback: The P, the I, and the DExplain what P, I, and D each respond to, the vice of each, and why tuning is a set of trade-offs.P acts on present error like a spring but always leaves a steady-state gap, because zero error means zero effort. I accumulates persistent error to close that gap, at the price of windup when the actuator saturates. D damps by responding to rate of change, at the price of amplifying sensor noise. Tuning trades speed against stability and accuracy against noise — and feedback that responds too late can excite rather than correct.
7. Kinematics: Forward Is Easy, Inverse Is HardContrast forward and inverse kinematics and explain workspace singularities.Forward kinematics chains joint transformations outward — one input, one answer, always. Inverse kinematics inverts every nice property: solutions may not exist, or a six-axis arm may have eight valid joint configurations for one tool pose. At singularities two axes align, a degree of freedom collapses, and required joint speeds head toward infinity — so arms use singularity avoidance and sometimes a redundant seventh joint.
8. The Loop Has a DeadlineExplain why real-time means guaranteed rather than fast, and why robots split fast and slow computation.In a control loop, late is identical to wrong: delay eats stability margin, because a correction applied late can push in the direction the error is already going. Real-time means the maximum delay is bounded and known — jitter is more damaging than latency, since a loop can be designed around a constant delay but not a variable one. Hence the split brain: a microcontroller running a 1–10 kHz loop that nothing may interrupt, and a Linux machine for planning and perception where a few extra milliseconds are harmless.
9. When the Loop Meets PeopleExplain how force sensing changes robot safety, and why a loop can only correct what it measures.A caged industrial arm has no proprioception of people — safety comes from geometry and from separate, simpler certified hardware, not from the robot's clever computer. A cobot adds force sensing (via joint torque sensors or motor current, since torque is proportional to current), making contact a measurable disturbance; the force and pressure limits, first issued as ISO/TS 15066 in 2016, were folded into ISO 10218-2:2025, which dropped the term "collaborative robot" for "collaborative application". But 'collaborative' is a property of the application, not the robot — and the deeper lesson is that a variable you never measured cannot be fixed in code.
10. What Makes It MechatronicsSynthesise the loop end-to-end and explain why mechanical, electrical, and software design cannot be separated.One command traverses everything: inverse kinematics picks joint angles, the encoder reports actual, PID converts error to current, the motor makes torque from that current, the gearbox makes it useful without adding backlash, and it all happens inside a deadline. Mechatronics exists because these domains are physically entangled — backlash destabilises control, encoder noise caps the D gain, and motor choice determines whether current can serve as a force sensor and therefore whether the robot can work near people.

Questions this course answers

Grounded in trusted sources

• Bolton, W. — Mechatronics: Electronic Control Systems in Mechanical and Electrical Engineering (Pearson)
• Craig, J. J. — Introduction to Robotics: Mechanics and Control (Pearson)
• Siciliano, B. & Khatib, O. (eds.) — Springer Handbook of Robotics
• Åström, K. J. & Murray, R. M. — Feedback Systems: An Introduction for Scientists and Engineers (Princeton University Press)
• Siegwart, R., Nourbakhsh, I. & Scaramuzza, D. — Introduction to Autonomous Mobile Robots (MIT Press)
• Thrun, S., Burgard, W. & Fox, D. — Probabilistic Robotics (MIT Press)
• Hughes, A. & Drury, B. — Electric Motors and Drives: Fundamentals, Types and Applications
• ISO 10218-1:2025 / ISO 10218-2:2025 — Robotics: safety requirements (Part 2 now carries the collaborative-application requirements formerly in ISO/TS 15066)

Every Wunder lesson is built from real, reputable sources — never invented.

Related Science courses