Noisy Measurements¶
Level: Elementary (Ages 8-12)
Why Can't We Just Measure Things Exactly?¶
Imagine you're trying to measure how tall you are with a ruler. Easy, right? But wait...
- Are you standing perfectly straight?
- Is the ruler perfectly straight?
- Are you reading it from exactly the right angle?
- Did you measure from the exact top of your head?
Every measurement has tiny errors. We call these errors noise.
What is Noise?¶
Noise is like static on a radio or fuzz on an old TV. It's the random errors that make measurements not quite perfect.
Example 1: The Bathroom Scale¶
Step on your bathroom scale 5 times in a row: - First time: 75.2 lbs - Second time: 75.4 lbs - Third time: 75.1 lbs - Fourth time: 75.3 lbs - Fifth time: 75.2 lbs
You didn't gain or lose weight between measurements! The scale just isn't perfect. That's measurement noise.
Example 2: Throwing Darts¶
Imagine you're really good at darts and always aim for the bullseye. But your throws land: - A little to the left - A little to the right - A little high - A little low - Right on target!
You're aiming at the same spot, but there's randomness in where the dart lands. That's noise!
Example 3: Temperature Sensor¶
You have a thermometer outside. You check it every minute: - 72°F - 73°F - 72°F - 71°F - 72°F
The temperature didn't really jump around that much! The thermometer has noise.
Types of Noise¶
Random Noise¶
Like throwing darts - sometimes high, sometimes low, but on average you hit the target.
Biased Noise¶
Like a scale that always reads 2 pounds too heavy.
Why Noise Matters¶
Problem 1: You Can't Trust One Measurement¶
If you measure your height once and get 4 feet 3 inches, is that exactly right? Maybe you're really 4 feet 2.8 inches, or 4 feet 3.2 inches!
Problem 2: Things Change¶
You're tracking a toy car rolling across the floor. You measure its position: - At 1 second: 10 inches - At 2 seconds: 19 inches (should be 20!) - At 3 seconds: 31 inches (should be 30!)
The noise makes it hard to know exactly where the car is.
Dealing with Noise¶
Strategy 1: Take Multiple Measurements¶
Instead of measuring once, measure many times and average them!
Measurements: 50, 52, 49, 51, 50
Average: 50.4
This is probably closer to the truth than any single measurement!
Strategy 2: Trust Your Prediction¶
If you know the toy car moves 10 inches per second: - At 2 seconds, you predict: 20 inches - You measure: 19 inches - You think: "It's probably between 19 and 20, maybe 19.5 inches"
You don't completely trust the noisy measurement!
Strategy 3: Use Multiple Sensors¶
Like having two thermometers: - Thermometer A says: 72°F - Thermometer B says: 74°F - You estimate: "Probably around 73°F"
Real World Examples¶
GPS in Your Phone¶
Your phone's GPS doesn't always show your exact location. Sometimes it shows you: - In the middle of the street (when you're on the sidewalk) - 20 feet away from where you really are - Jumping around even when you're standing still
That's GPS noise! Your phone uses clever tricks (like the Kalman filter!) to figure out where you really are.
Video Game Controllers¶
When you hold a game controller still, it might detect tiny movements. That's noise from the sensors! Games filter this out so your character doesn't jitter.
Robot Vacuum Cleaners¶
A robot vacuum has sensors to detect walls and furniture. But the sensors aren't perfect - sometimes they see things that aren't there, or miss things that are! The robot has to be smart about what to believe.
The Big Idea¶
We can't measure things perfectly, so we need to be smart about: 1. Not trusting any single measurement completely 2. Combining multiple measurements 3. Using predictions to help 4. Filtering out the noise
This is exactly what the Kalman filter does!
Try It Yourself!¶
Experiment 1: Noisy Ruler¶
- Draw a line exactly 10 cm long (use a ruler carefully)
- Have 5 friends measure it with their own rulers
- Write down all the measurements
- Are they all exactly 10 cm? Probably not!
- Calculate the average - is it close to 10 cm?
Experiment 2: Reaction Time¶
- Use an online reaction time test
- Take the test 10 times
- Write down all your times
- Notice how they're all different? That's noise!
- What's your average reaction time?
Experiment 3: Counting Steps¶
- Walk 20 steps normally
- Have a friend count your steps
- Do it again - have another friend count
- Do it a third time - have a third friend count
- Did they all count exactly 20? Probably not!
Experiment 4: Temperature Tracking¶
- Check a weather app every hour for a day
- Write down the temperature
- Make a graph
- Does it change smoothly, or jump around?
- Some of those jumps are noise!
Questions to Think About¶
- Why do you think scales show different numbers when you step on them multiple times?
- If you measure something 10 times and get 10 different answers, which one is "right"?
- Is it better to trust one very careful measurement, or the average of 10 quick measurements?
- Can you think of other things in your life that have "noise"?
What's Next?¶
Now that we know measurements are noisy, the next chapter will teach us how to combine different pieces of information to make better estimates!
Key Vocabulary - Noise: Random errors in measurements - Measurement: Using a sensor or tool to find out something - Average: Adding up numbers and dividing by how many there are - Sensor: A device that measures something (thermometer, scale, GPS, etc.) - Bias: When measurements are consistently wrong in the same direction