Using data to improve robots

Learn

What you’ll learn

• What to measure and how to log it
• How to run a fair test
• How to turn numbers into decisions

Measure, don’t guess

“It feels faster” isn’t evidence. Measure:

• Time (lap time, time to goal)
• Accuracy (distance from target, times it lost the line)
• Reliability (successes out of 10 runs)
• Energy (battery drop per run)

Logging

Print data with a timestamp so you can copy it into a spreadsheet:

if (millis() % 500 < 20) print(millis() + "," + linePosition());

On a real Arduino, Serial.println() the values and copy them from the Serial Monitor, or use the Serial Plotter to see graphs live.

Fair tests

• Change one variable at a time.
• Keep everything else the same (track, battery level, start position).
• Repeat each test (3–5 runs) and use the average.
• Write it down straight away.

Decide with evidence

Compare averages in a table, and pick the setting that best meets your requirements, not just the fastest one. A robot that’s 10% faster but fails 3 times out of 10 usually isn’t better.

Examples

Example: a tuning table

KP	Run 1	Run 2	Run 3	Average	Lost line?
20	41 s	42 s	40 s	41 s	0/3
40	33 s	34 s	33 s	33.3 s	0/3
80	30 s	—	31 s	—	1/3

Decision: KP = 40. KP = 80 is a little faster but unreliable.

Example: a lap timer in the simulator

let lapStart = 0, away = false;
function loop() {
  // the start is on the top straight; linePosition is near 0 there
  const t = millis();
  if (t - lapStart > 5000) away = true;
  // ...your controller here...
}

Practice

Try it

Log while you tune

▶ Run↺ Reset

Your code (JavaScript)const BASE = 50, KP = 40; function loop() { const e = linePosition(); if (e === null) { motors(30, -30); return; } if (millis() % 1000 < 20) print(millis() + " ms, error " + e); motors(BASE + KP * e, BASE - KP * e); }Load starter code

Robot commands

• motors(left, right): wheel power, −100 to 100
• lineLeft(), lineRight(): true over the black line
• distance(): cm to the nearest thing ahead (max 200)
• distanceLeft(), distanceRight(): cm to the sides
• linePosition(): where the line is, −1 (left) to 1 (right), or null
• millis(): milliseconds since Run
• print(value): show a message below
• setup() runs once; loop() runs 50 times a second

Average three runsEasy

Time three laps (watch the time under the arena) at BASE = 50 and work out the average.

Fair testMedium

Test BASE = 40, 55 and 70 with KP fixed, three runs each. Fill in a table and choose the best setting.

Justify itHard

Write two sentences recommending a final setting to your team, using your numbers as evidence.

Quick check

1. A fair test changes…

Everything at onceOne variable at a timeNothingOnly the code comments

2. Why repeat each test several times?

To average out random variationTo drain the batteryIt’s not neededTo make it slower

3. The fastest setting fails 3 out of 10 runs. Is it best?

AlwaysUsually not, if reliability mattersYes, speed is all that countsOnly on Tuesdays

Check my answers