Koinder Practitioner Track 18+ ← My Learning About the track Clubhouse

Robotics: The Physical Craft

Not a branch of AI, and two centuries older. Mechanical engineering, electronics and control theory — where sensors lie, motors wear out, and failure can hurt somebody.

Module 10 · Practitioner Track · 18+

The one thing almost everybody gets wrong

Robotics is not a branch of artificial intelligence. It is a separate engineering discipline, and it is older.

Robotics is mechanical engineering, plus electronics, plus control theory. Its subject is machines that physically sense and act on the world. Whether any intelligence is involved is a separate question, and usually the answer is no.

Robots with no AI at all The Unimate arm, 1961 — repeated one taught sequence, forever. A CNC milling machine. A 3D printer. A warehouse vehicle following magnetic tape. An automatic gate. Most industrial robots working in the world today.
AI with no robot at all Claude. A spam filter. A credit-scoring model. A chess engine. A recommendation system. No sensors, no motors, no physical presence anywhere.
The da Vinci surgical robot

Four arms, operating inside a human body, extraordinary precision. It makes no decisions whatsoever. A surgeon controls every movement in real time. It is a very sophisticated robot with zero autonomy — and that is a deliberate choice, not a limitation.

Zipline, Rwanda and Ghana

Drones delivering blood and vaccines to rural clinics, flying beyond the operator's sight, and dropping payloads by parachute. Africa leads the world in medical drone delivery, and the difficulty is overwhelmingly aerodynamics, batteries, wind and regulation — not intelligence.

A different ancestor entirely

AI descends from logic. Robotics descends from control engineering, and it is two centuries older than computing.

1788 — Watt's centrifugal governor

Two spinning balls on a steam engine. Faster spin, balls rise, steam valve closes, engine slows. A complete feedback control loop, built from brass, 160 years before the first computer. Every robot alive is a descendant of this.

1868 — Maxwell, "On Governors"

The mathematics of why some of these loops settle and others shake themselves apart. This is where control theory becomes a science.

1960 — Kálmán

How to estimate truth from measurements you cannot trust. The Kalman filter flew on Apollo, and it is in the phone in your hand right now.

1961 — the Unimate

Engelberger's arm at General Motors, lifting red-hot metal so that men did not have to. The point was never intelligence. It was hands.

Why robotics is harder than it looks — five things software people learn painfully

1. Sensors lie, constantly
The ultrasonic distance sensor

Costs about ₦1,500 and returns a number in centimetres. It also returns nonsense off soft fabric, at an angle, in wind, near another sensor, or when the temperature changes. Your code must expect a wrong reading, because it is coming.

2. The world is analog and does not repeat
"Turn the motor for 2 seconds"

Ships different distances on tile and on carpet. Different again when the battery is at 80% versus 40%. Different when the motor is warm. This is why robots need feedback rather than instructions — the world will not hold still for you.

3. Everything wears out
Backlash

A new gearbox has a tiny amount of play. Six months later it has more. Your carefully calibrated arm now misses by three millimetres and no code changed. Mechanical drift is a maintenance schedule, not a bug.

4. Power is a hard budget
The drone tradeoff

More battery means more flight time and more weight, which needs more power to lift, which needs more battery. Every gram is negotiated. Software engineers are used to "add more" being free. Here it is never free.

5. Failure hurts people
The stakes are physical

A web bug shows an error page. A robotics bug breaks a bone. This is why industrial robots have physical emergency stops wired to cut power directly — circuits that cannot be overridden by software, because software is what might have failed.

If you take one habit from this module: design the stop before you design the motion.

Where robotics genuinely pays in Nigeria

Ignore humanoid robots. The money is in unglamorous machines that solve an actual local problem.

Irrigation control

Soil moisture sensor, a valve, a controller. Waters only when the soil is dry, not on a timer. Saves water and diesel. Parts cost under ₦40,000 and it is a genuine sense–think–act loop.

Poultry house environment

Temperature and ammonia sensors driving fans and misters. Birds are extremely sensitive and a bad afternoon can cost a farmer a great deal. Straightforward control, real value.

Generator automation

Detect mains loss, wait, start the generator, transfer load, reverse when mains returns. Every serious building here needs this. Notice it needs no AI and considerable safety engineering.

Cold chain monitoring

Temperature logging on vaccine and produce transport, with alerts. Sensing and reporting, no actuation. Still robotics work, and it saves lives.

What a robotics engineer actually does on a Tuesday

  • Works out why the arm is 4mm off and whether it is code, calibration or a worn gear
  • Reads a datasheet to find out what a sensor does at 38°C
  • Tunes a control loop by hand because the model does not match the real machine
  • Argues for a mechanical fix instead of a software workaround
  • Tests the emergency stop, again

Practical exercise. Pick one repetitive physical task near you. Write down exactly what it must sense, what decides, what moves — and then what must happen if the sensor fails. Most people skip that last one, and it is the part that keeps somebody's hand attached.

🤖

One question before we start

The Practitioner track is built for adults. It assumes you are ready to deploy real systems that real strangers will use, and to be answerable for what they do.

We ask because in Nigeria you become an adult at 18, and this track is a paid commitment. Nothing here is stored for anyone under that age.