What is a robot, actually?
Every robot ever built does three things. Once you see the pattern, you can't unsee it.
The robot in your head β the one from movies β probably walks, talks, and has opinions about humanity. That robot doesn't really exist yet.
The ones that do exist are stranger. And honestly more interesting.
Every robot does three things
No matter how simple or complex, every robot ever built follows the same loop:
Sense β It takes in information from the world. A camera. A microphone. A distance sensor. GPS coordinates. The robot's version of eyes and ears.
Think β It processes that information and makes a decision. What does this data mean? What should I do next?
Act β It does something physical in the world. Moves a wheel. Opens a valve. Picks up an object. Turns a camera.
This is called the Sense-Think-Act loop. It's the skeleton of every robot ever built β from a Β£20 toy to a Mars rover.
Three robots. Same loop.
A Roomba (robot vacuum)
- Sense: Bump sensors detect walls. Cliff sensors detect stairs.
- Think: "I hit something. Turn right and keep going."
- Act: Wheels reverse, robot turns, continues cleaning.
The Roomba's "Think" step is a simple rule. No learning, no memory. Just: if this, then that.
A factory robotic arm
- Sense: Position encoders track the arm's exact angle at every joint.
- Think: Follow the pre-programmed path to the next waypoint.
- Act: Motors move each joint with millimetre precision.
The arm is fast and accurate β but only because its world is completely controlled. Move one bolt on the assembly line and it's lost.
A self-driving car
- Sense: Cameras, radar, and LiDAR build a 3D map of everything nearby.
- Think: Neural networks identify objects; path-planning algorithms decide the safest route.
- Act: The car steers, accelerates, and brakes β without a human hand on the wheel.
The self-driving car's "Think" step is orders of magnitude more complex than the Roomba's. But the loop is identical.
Why this matters
Once you know the Sense-Think-Act loop, you start seeing it everywhere β in robots, yes, but also in thermostats, autopilots, and recommendation algorithms.
It also tells you exactly where robots fail. A Roomba gets confused when its sensors are dirty. A factory arm stops working when the environment changes. A self-driving car struggles in heavy rain when its sensors are degraded.
Every robot failure is a failure in one of three places: it sensed wrong, it thought wrong, or it acted wrong. That's the whole list.
Want to go deeper on any part of the loop? The Atlas has entries on sensors, actuators, and control systems.
Check your understanding
These aren't graded. Just think about them.
1. A motion-sensor light turns on when you walk past. Which part of the Sense-Think-Act loop is the motion sensor? Which part is the light turning on?
2. A Roomba bumps into a wall and turns. Where in the loop did something go "wrong" β or did anything go wrong at all?
Here's something worth sitting with: a thermostat senses temperature and decides to turn on your heating. By the definition above β is it a robot?
Community discussion
0 questions & insightsLoading discussionβ¦
Spotted something off? Report an error β