Why Uncertainty Matters

Robots Live with Imperfect Information

What if the hardest part of robotics is not building the robot, but dealing with imperfect information?

Every real robot works with data that is a little bit wrong. Its sensors, motors, and even its own internal math are never perfectly accurate.

This is called uncertainty, and it is everywhere in the real world.

Learning where uncertainty comes from is the first step toward building robots that work reliably outside the lab.

Five Sources of Uncertainty

Real-world robots face uncertainty from five primary sources.

Sensor noise means readings wobble around the true value. Actuator imprecision means motors and wheels do not move exactly as commanded.

Model inaccuracies mean the robot's internal picture of itself and the world is simplified. Computational limitations mean it cannot calculate everything perfectly in time.

Environmental unpredictability means the world changes in ways nobody planned for.

Example: Wheel Slip in a Hallway

Picture a mobile robot driving down a hallway.

Its wheel encoders count wheel turns and report that it traveled 10 meters.

But wheels can slip on the floor. Because of wheel slip, the actual distance might be anywhere from 9.5 to 10.5 meters.

Notice that the reported position is a single point, while the robot could really be anywhere inside the shaded range around it.

The reported number looks exact. The true position is really a range of possibilities.

How Small Errors Pile Up

A deterministic system trusts the reported values as if they were exact.

So the robot believes it moved exactly 10 meters, every time.

Each move adds a small hidden error. The next move starts from that wrong spot and adds another error.

Soon the errors accumulate, and the robot loses track of where it is.

Deterministic vs Probabilistic Thinking

Traditional deterministic approaches fail because they cannot represent or reason about uncertainty. They hold one answer and treat it as the truth.

Probabilistic approaches keep track of a range of possibilities and how likely each one is.

For the hallway robot, that means thinking "somewhere between 9.5 and 10.5 meters" instead of "exactly 10."

Admitting doubt makes the robot more robust, not weaker.

Putting It to Work

When a robot behaves strangely, ask which of the five sources is at play.

Is a sensor noisy? Did a motor or wheel not move as commanded? Is the model too simple, the computer too slow, or the world surprising?

Then check how the system handles it. If it trusts every reading as exact, it is deterministic and will drift.

Reliable robots treat information as uncertain and reason with probabilities.

A deterministic robot system fails under uncertainty mainly because it _____

• cannot represent or reason about uncertainty
• runs too slowly on robot hardware

Answer: cannot represent or reason about uncertainty

Deterministic approaches treat reported values as exact truth. They have no way to represent or reason about uncertainty, so errors go unnoticed and uncorrected.

In robotics, _____ corrupts what the robot perceives, _____ means commanded motions are not executed exactly, and _____ comes from changes in the world around the robot.

Answer: In robotics, sensor noise corrupts what the robot perceives, actuator imprecision means commanded motions are not executed exactly, and environmental unpredictability comes from changes in the world around the robot.

Sensor noise affects perception, actuator imprecision affects motion execution, and environmental unpredictability comes from a changing world. Together with model inaccuracies and computational limitations, these are the five primary sources of uncertainty.

Which of these is NOT one of the five primary sources of uncertainty for real-world robots?

• Programming language choice
• Model inaccuracies
• Computational limitations
• Sensor noise

Answer: Programming language choice

The five sources are sensor noise, actuator imprecision, model inaccuracies, computational limitations, and environmental unpredictability. Programming language choice is not one of them.