# Week 6 — Encoders, calibration, and repeatability

DIS OpenLab · Foundations of Autonomous Systems · v0.1.0-prepilot · 2026-10-15
Track: disturbed computational model and supplied log schema; physical work only with an approved reference station
License proposal: CC BY 4.0 for original text, pending rights review

## Learning objectives

Learners will convert encoder counts to wheel angle; plan PWM-to-speed and geometry calibration; repeat trials; summarize spread instead of selecting the best run; record configuration and uncertainty; and identify when a component change invalidates prior calibration.

## Core calculations

For `N` counts per wheel revolution and a measured count change `Δc`:

```text
wheel angle rad = 2π·Δc / N
distance m = wheel radius m · wheel angle rad
```

Confirm whether the encoder specification is per motor revolution or wheel/output-shaft revolution and whether quadrature edges are already multiplied. Gear ratio and missed counts can dominate error.

## Safety boundary

Physical calibration requires a guarded low-speed area, reachable manual isolation, secure wheels/mounts, current-limited supply, and adult supervision. Stop for heat, odor, unexpected acceleration, damaged wiring, motor stall, loose hardware, or loss of control updates.

## Session A — computational wheel mismatch

Start from a straight command. Model the right effective wheel radius as 2%, then 5%, smaller than the left. Predict heading drift after a fixed duration. Keep wheelbase and time step unchanged.

Record:

| Run | Left scale | Right scale | Duration s | Final x m | Final y m | Heading rad |
|---|---:|---:|---:|---:|---:|---:|
| Nominal | 1.00 | 1.00 | | | | |
| Mismatch A | 1.00 | 0.98 | | | | |
| Mismatch B | 1.00 | 0.95 | | | | |

State whether the result is calculation, deterministic model, native Webots, or physical observation.

## Session B — calibration plan

Design a procedure before collecting data:

1. identify the exact configuration and revision;
2. mark a fixed start pose and test distance;
3. select a low command and duration;
4. execute at least three comparable trials;
5. record every trial, including failures;
6. compute mean and range (or standard deviation if taught);
7. change only one factor, then repeat;
8. define acceptance criteria before viewing results.

Use `calibration-log.csv`. Leave measurement fields blank when no physical run exists.

## Encoder and command interface table

| Field | Value | Evidence/source |
|---|---:|---|
| Encoder counts/revolution definition | | |
| Gear ratio | | |
| Wheel radius m | | |
| Wheelbase m | | |
| PWM frequency Hz | | |
| Command deadband | | |
| Maximum approved command | | |
| Watchdog/stop threshold s | | |

## Evidence to submit

- three-run computational mismatch table;
- complete proposed calibration procedure;
- configuration table with unknowns left explicit;
- blank or genuinely measured calibration log, correctly labeled;
- explanation of repeatability versus accuracy;
- change-control note identifying which substitutions require recalibration.

## Instructor notes

Do not accept a single successful path as repeatability. If teams only have supplied logs, ask them to detect outliers without deleting them. “Calibration factor” should name input, output, units, operating range, and revision.

A chassis substitution can alter wheelbase; a tyre change can alter effective radius and traction; a battery/supply change can alter motor response; a firmware timer change can alter PWM. Treat these as configuration changes.

## Low-bandwidth and accessibility option

Provide three printed trial records. Learners can calculate mean/range with a basic calculator and annotate the table. Assign measurement, recorder, safety observer, and analyst roles so fine-motor tasks are not a participation gate.

## Exit ticket

Explain why “the rover travelled one metre once” is not a repeatability claim, and state the minimum additional records needed.
