RL Robotics LabSearch ↗
Lesson / authored

Week 13 · Improve: Calibration validity

Session 4 of 4 · Read the world · Phase 4

Plan about 15 minutes for explanation, 30 minutes for practical work and 10–15 minutes for documentation. A longer build may continue into the next session: stop safely, commit the current state and record the next check. Desktop simulations count as software evidence; label them clearly and record physical validation separately.

Engineering challenge

What does a sensor number mean when the surface, height or lighting changes? This session focuses on calibration validity.

Before you start

The previous week’s recorded baseline and Week 12, Improve. For later sessions this week, retain the preceding session’s files and predictions.

Equipment: Desktop Python, editor, paper and ruler; for physical work, the configured 3pi+ 2040, clear floor mat and hardware checklist. Week 3 additionally uses the separate low-voltage LED circuit described in its procedure.

For any motion, verify the stop button, short time limit and clear floor area first. Keep the wheels raised for a new device program until its commands and stop behaviour are checked. A hazard or uncertain input is a reason to stop and document, not to force the trial to finish.

Theory and mathematics

Calibration validity

Calibration should span both conditions a detector must distinguish and the environmental variation expected in use. If dark and light reading ranges overlap, no single threshold can classify every example perfectly. Adding more samples from the same easy condition does not fix missing difficult cases. Reserve new cards or positions for validation. For bump sensors, calibrate with the skirt released and test left/right contact separately with motors off. Unknown or stale readings should trigger a conservative decision; their absence is not evidence that a route is clear.

Worked example — illustrative values

For fictional light readings 120, 130 and 125 and dark readings 800, 820 and 810, means are 125 and 810 raw units. A midpoint candidate is 467.5. It separates these six examples, but you still need independent readings at new positions before trusting it.

Write the calculation in your notebook before running code. State which values you measured, which you assumed and which the program calculates. A correct numerical calculation cannot rescue an incorrect physical assumption.

Run and explain the model

The following is desktop Python, not a ready-to-run motor program. Download this week’s example, save it in your student repository and run python3 code/w13.py from the repository root. The same small model is reused across the week so you can learn it, build with it, test it and revise it.

# Desktop Python teaching example. Numerical inputs are illustrative.
from statistics import mean
light = [120, 130, 125]  # fictional raw units
dark = [800, 820, 810]
threshold = (mean(light) + mean(dark)) / 2
for reading in [125, 470, 815]:
    print(reading, "dark" if reading > threshold else "light")

Compare the unchanged baseline and your proposed change on the same inputs. Keep the original files so another reader can reproduce the comparison. If an exception appears, read its final line, identify the input or assumption that caused it and make the smallest explained correction. Do not delete validation merely to obtain output.

Understanding the model and its limits

The threshold is the midpoint between two labelled means in raw sensor units. It is a simple baseline, not necessarily the best decision boundary. Real light/dark distributions can overlap, and the five channels may have different scales. Preserve readings by channel and condition rather than averaging away a weak sensor. The illustrative reading 470 lies near the decision boundary, so small variation may change its class. Collect independent labels and test repeated readings there. Reflectance classification describes a floor surface; it does not directly say a free-standing obstacle is a certain distance away.

Practical instructions

  1. Select a calibration dataset spanning the intended floor fixtures.
  2. Reserve ten readings at new positions as validation data.
  3. Document overlap, channel differences and any condition requiring a conservative stop.
  4. Store calibration date, raw units and fixture description for Week 14.

Experiment

Test the frozen calibration on ten held-out readings. State whether the original labels remain separable under the chosen rule.

Before testing, record your prediction, changed factor, measured response, fixed conditions and stopping rule. Save every attempted run, including failures, with a condition and source version. If hardware is unavailable, use an explicitly labelled synthetic/replay dataset and list the physical question it cannot answer. Do not invent completed trials.

Deliverable

A calibration with held-out evidence, validity conditions and an unknown-reading policy.

Save notebook/w13-s4.md, the relevant code revision, raw CSV or test-case records, and one labelled diagram/plot/table. Link the files relatively from your notebook. Use the entry template and report guide.

Completion criteria

A documented failed prediction can meet the learning criteria. A missing physical trial must remain marked untested; software success alone does not validate the robot.

Reading and video

Reflection and next step

Which assumption most affected your result? Point to one observation that supports your explanation and one alternative explanation the evidence has not ruled out. Write a specific next test with a changed factor and measurable outcome, then proceed through the week’s Learn → Build → Experiment → Improve cycle.

← Previous sessionCourse roadmapNext session →