What a Robot Can Feel: Inner Senses vs Outer Senses

A Trick You Already Know

Close your eyes and touch your nose. You did it without looking, because some part of you always knows where your own arm is — whether the elbow is bent, how far the hand has swung. That inner body-sense is called proprioception. Now open your eyes and reach for a coffee cup across the table. This time you need your eyes: nothing inside your body tells you where the cup is — only your senses pointed at the outside world can. That outward sense is called exteroception.

A robot's senses split along the exact same line. This single division — am I measuring my own body, or am I measuring the world? — is the most useful map of the whole sensing chapter. It is named proprioceptive vs exteroceptive sensing, and once you see it, every sensor you meet later drops neatly into one bucket or the other.

Inner Senses: Knowing Your Own Body

Proprioceptive sensors measure the robot's own state — how its joints are positioned, how fast it is turning, how hard a motor is pushing. They are usually small, cheap, fast, and built right into the joints and electronics. Three show up almost everywhere.

1. A rotary encoder sits on a motor shaft and counts how far the joint has turned — this is the robot's sense of where its limbs are, the direct equivalent of knowing your elbow is bent.
2. An inertial measurement unit (IMU) bundles an accelerometer (which feels pushes and the constant tug of gravity) with a gyroscope (which feels turning). Together they tell the robot which way is up and how its body is tilting — your own inner-ear balance, in a chip.
3. A force/torque sensor at the wrist, or current sensing inside the motor driver, tells the robot how hard it is pushing or being pushed — its sense of effort and contact, like feeling the weight of a bag in your hand.

The beauty of inner senses is that they keep working in total darkness, dense fog, or empty space — they never depend on the world cooperating. Their weakness is that they only describe the body, never the room around it. An encoder will happily tell you a wheel spun ten times even if the wheel was spinning uselessly on ice.

Outer Senses: Reading the World

Exteroceptive sensors point outward and measure things the robot does not control: where the walls are, what object is on the table, where the robot sits on a map. They are how a robot finds the cup. The common ones each have a personality.

1. A LiDAR sweeps laser beams around and times how long each one takes to bounce back, drawing a crisp ring of distances — a robot feeling out the room with thousands of invisible tape measures per second.
2. An RGB-D / depth camera gives a normal color picture plus a distance for every pixel, so the robot sees both what an object looks like and how far away it is.
3. An ultrasonic (sonar) sensor chirps sound and listens for the echo — cheap, short-range, and the same trick a bat uses to fly in the dark.
4. A GNSS / GPS receiver listens to satellites overhead to fix the robot's place on the whole planet — an outdoor sense of where in the world you are, accurate to a few meters.

Outer senses are powerful but moody. Cameras struggle in glare or darkness, LiDAR is confused by glass and rain, sonar gets fooled by soft surfaces, and GPS vanishes indoors or between tall buildings. They depend on the world cooperating — which is exactly why a robot can never trust just one.

Why a Robot Needs Both, and What Happens Next

Inner and outer senses cover for each other. An IMU keeps a drone level for a fraction of a second even when the camera is blinded by the sun; a camera corrects the slow drift that any IMU accumulates over minutes. Blending several imperfect readings into one better estimate is called sensor fusion, and it is the quiet engine behind almost every capable robot.

All of this feeds the most basic loop in robotics: sense, then plan, then act, then sense again. That loop is the sense–plan–act paradigm. Sensing is step one, so the quality of every later decision is capped by how well the robot felt the situation in the first place. Garbage in, garbage out — which is why later guides spend so long on raw-signal problems like noise, drift, and calibration.

A Map of the Whole Chapter

Here is the rest of this chapter sorted by the one question you now carry with you. Keep it nearby: as each later guide introduces a sensor, ask yourself which column it belongs in, and the new idea will already have a home.

QUESTION THE SENSOR ANSWERS        SENSORS YOU'LL MEET
---------------------------------  -----------------------------
INNER  "What is my body doing?"    encoder, IMU (accel + gyro),
  (proprioceptive)                 force/torque, tactile

OUTER  "What is the world doing?"  LiDAR, RGB-D camera, sonar,
  (exteroceptive)                  GNSS / GPS

BOTH   "How much do I trust it?"   noise, bias & drift,
  (applies to every sensor)        calibration, sampling,
                                   filtering, sensor fusion

Notice the third row. Noise, calibration, and fusion are not a kind of sensor at all — they are the rules for handling whatever any sensor hands you. That is why the chapter ends there: once you know what a robot can feel and how much to believe it, you are ready to turn raw signals into real understanding.