Sensors, microprocessors and actuators — the loop that runs every automated system. Then robots (what they are, where they're used, who they replace), artificial intelligence, machine learning, and expert systems. This is the topic examiners can't stop asking about — every year, in some form.
Every automated system is built on three parts: a sensor reads the environment, a microprocessor decides what to do, and an actuator carries it out — looping continuously until the system is switched off. A robot is a portable automated system in a mechanical body: mechanical structure + electrical components + programmable. Artificial intelligence is any program that simulates human behaviour (collect data, use rules, reason, sometimes learn). Machine learning is the type of AI that adapts its own rules. An expert system is the type of AI that runs a human expert's knowledge as facts + rules + an inference engine. Lock those four ideas in and the whole topic opens up.
| Objective | Why it matters in the exam |
|---|---|
| Describe how an automated system uses sensors, a microprocessor and actuators | The 6-mark "explain how the sensor and microprocessor are used" question appears almost every year — the marks live in the 6-step chain. |
| Explain the advantages and disadvantages of automation | Watch who benefits — an advantage "to the worker" is not one "to the employer" (2025 examiner report). |
| Describe the three characteristics of a robot | Mechanical · electrical · programmable. AI is NOT one of them — a common lost mark. |
| Describe uses of robots in context (six domains) | One benefit + one drawback in context beats a generic list. |
| Describe AI, machine learning and expert systems | Examiners see these three mixed up every sitting — they are nested, not identical. |
| # | Step |
|---|---|
| 1 | The sensor reads/measures the environment (temperature, distance, motion, light…). |
| 2 | The analogue reading is converted to digital by an ADC. |
| 3 | The digital data is continuously sent to the microprocessor. |
| 4 | The microprocessor compares the data to a stored value / range. |
| 5 | If the condition is met, it sends a signal to the actuator. |
| 6 | The actuator performs the action. The process repeats until the system is switched off. |
① Mechanical structure / framework · ② electrical components (sensors, microprocessor, actuators) · ③ programmable / can move. AI is not a characteristic.
① collection of data · ② set of rules · ③ ability to reason · ④ ability to learn / adapt. Any three earns the marks.
Knowledge base (facts) · rule base (links facts) · inference engine (picks next question / fires the rule) · interface (user enters answers, reads diagnosis).
Supervised: a person labels the training data. Unsupervised: no labels — the program finds patterns itself, usually by clustering.
Every automated system is built on the same three parts: a sensor to read the environment, a microprocessor to decide what to do, and an actuator to actually do it. Robots are essentially portable automated systems, wrapped in a mechanical body and programmed to move. Artificial intelligence goes further — instead of following fixed rules, an AI program can reason and, in many cases, learn. Machine learning is the specific sub-field where the program can change its own rules. And an expert system is a very old, very specific type of AI: it stores a human expert's knowledge as facts + rules, and uses an inference engine to walk through them until it reaches an answer. Get those four ideas locked in and you can answer almost anything in this topic.
Pick a device you've used today that had a sensor deciding what to do — an automatic tap, a lift, a phone screen dimming, an auto-door, a car headlight coming on. For that device, name: (1) the sensor and what it reads, (2) the decision the microprocessor makes, (3) the actuator that moves. If you can't name all three, you've found where to focus.
Modern tractors now run a system called See & Spray that combines every part of Topic 6 in one machine. Cameras (sensors) capture 20 images per second of the ground. An onboard AI (using machine learning trained on millions of images) tells weed from crop. If the pixel is a weed, a microprocessor fires a signal to a nozzle (actuator) that sprays herbicide only on that spot. Result: up to 90% less herbicide, at highway speed. It's an automated system, running a robot, powered by AI. Every exam question about "explain how AI/robots/automation are used" is really asking you to describe something like this.
If a self-driving car has to choose between hitting a pedestrian or swerving into a wall (harming its passenger), who chose the rule? The programmer? The car company? Is that intelligence, or is the car just running a pre-coded rule that a human wrote years earlier? Mark schemes here lean towards a very careful answer: the machine follows the rules, the humans wrote the rules, and machine learning changes the rules based on data. Never write "the AI decided" as if AI has intent — write "the program applied the rule".
Three concepts, examiners see them mixed up every single sitting. They are nested — machine learning is a type of AI, expert systems are also a type of AI. But they're not the same thing.
The broad field. Any program that simulates human behaviour — collecting data, using rules, reasoning, and sometimes learning. AI can include everything from a chess opponent to Siri to an expert system.
Four features: data · rules · reason · learn/adapt.
A specific type of AI where the program can change its own rules or data based on results. Supervised (a person labels the training data) or unsupervised (the program groups data itself).
Key phrase: "adapts its stored rules or processes".
A specific type of AI that mimics a human expert. Four parts: knowledge base (facts), rule base (links between facts), inference engine (decides which question next), interface (talks to user).
Doesn't have to learn — an expert system with fixed rules still counts.
Pick a scenario. Drag the environment slider to change the real-world condition. Then press Run one cycle and watch the loop fire: sensor reads → ADC digitises → microprocessor compares to a stored value → decision is made → actuator either fires or stays idle. This is the exact chain Cambridge asks you to describe in the "explain how the sensor and microprocessor work together" question that shows up every year.
This is the highest-yield question in Topic 6.1 — it comes up almost every year and the mark scheme wants a chain of 6 specific points. Learn the chain. Adapt the wording to the scenario in the question.
Examiner report says: weaker candidates give a "generic description" of a sensor sending data to a microprocessor with no reference to the given context. The context is where marks live — say "the temperature sensor measures the greenhouse temperature", not just "the sensor reads data".
A tiny expert system that diagnoses why your printer isn't working. Answer each question and watch three things happen simultaneously: facts get added to the knowledge base, rules fire in the rule base, and the inference engine works out which question to ask next. When it reaches a conclusion, it explains itself.
Is the green power light on the printer illuminated?
Examiners want you to match a robot to a domain and give one benefit + one drawback in context. Learn one benefit and one drawback per domain — you'll rarely need more.
| Domain | Example | Benefit (mark-scheme wording) | Drawback (mark-scheme wording) |
|---|---|---|---|
| 🏥 Medicine | Surgical robot, nurse-bot, robot pharmacist | Surgeon can operate from anywhere; more precise than a human hand | If the network drops the procedure stops; expensive; programming errors could harm the patient |
| 🌾 Agriculture | Automated tractor, seed-planter, robot herder | Works 24/7; frees the farmer for other tasks; more consistent than a human | High initial cost; maintenance required; can replace farm workers |
| 🚗 Transport | Self-driving car, warehouse mover, delivery drone | Faster reactions than a human = safer; no fatigue | Expensive; cyber-security risk if someone changes the program; job losses |
| 🏭 Industry | Car assembly robot, packing robot, precision welder | Consistent quality; no breaks; more precise than humans | High set-up cost; replaces workers; needs specialist maintenance |
| 🎮 Entertainment | Robot dog toy, drone camera, educational robot kit | Fun / engages people with technology; teaches programming | Battery-limited; connection loss can cause a drone crash |
| 🏠 Domestic | Robot vacuum, robot lawnmower, security patrol bot | Saves time on repetitive tasks; can run when you're out | Cannot handle stairs; still needs monitoring; upfront cost |
The 2025 mark scheme accepted any three of these for a 3-mark "characteristics of AI" question. Learn all four so you always have three to write down.
The program needs data to work on — from the user, or from its sensors.
Stored rules that the program uses to make decisions from the data.
Uses logic: if rules + facts, then a conclusion. E.g. all dogs eat meat + Fred is a dog → Fred eats meat.
Change its own rules or data (machine learning). Not every AI does this — only the ML-flavoured ones.
Stores facts from the human expert (e.g. "green light means power on").
Stores rules linking the facts (e.g. "IF green light off THEN no power").
Decides which question to ask next. Uses previous answers to pick a path through the rules.
Where the user enters answers and reads the diagnosis. Input + output.
A person labels the training data. E.g. "this image is a dog", "this image is a horse". The program learns the features of each label. Then when it sees a new image, it tries to classify it.
Key phrase: "the user tells the program what the data means".
No labels. The program is given data and finds patterns on its own — usually by clustering (grouping items that are close together in some feature space).
Key phrase: "the program learns from the data without human input".
A device is described. Say whether the highlighted part is a sensor, a microprocessor, or an actuator.
Which sensor best fits which task? Click one on the left and its matching partner on the right.
Which domain does this robot belong to?
Click the six steps of the automation loop in the correct order.
Rapid-fire mixed questions. Score as many as you can in 60 seconds.
Real question style from past papers (May/June 2023–2025 and specimen). Write your answer, then hit Check to see how many mark-scheme keywords you hit.
One badge per skill in Topic 6.1–6.4. Green = secured (≥75% correct or manually marked). Yellow = focus (attempted but not secured). Grey = not attempted. Tap any badge to override.