Running a session

A shape that holds up

0–10 min	Everyone signs in and opens their dashboard. You walk the room and read where the group is. Arms powered, connected, twin following.
10–20 min	You demonstrate the one thing in today's module that is hard to discover — the tool calibration, the ID assignment, the HSV sliders. Once, at the front, with the projector.
20–75 min	Pairs work. You circulate.
75–85 min	Someone shows the room what worked. Two minutes each, three pairs.
85–90 min	Power off, cables coiled, arms parked in a safe pose.

The demonstration slot matters more than it looks. The lessons are good at telling a learner what to do and less good at showing a fourteen-year-old how the four-point tool calibration is physically performed.

Connecting, every time

The first five minutes of every session are the same four checks, and it is worth making them a ritual the class does without prompting:

1. Power on, then USB in.
2. Connect in the Lab, and read the motor table — six OK.
3. Move a joint by hand and watch the twin follow.
4. Confirm the emergency stop is on screen.

A pair that skips step 2 and starts a lesson with five motors will lose twenty minutes to a mystery.

Rotating through one arm

With one arm per pair or per four, run the module in two lanes:

• Virtual lane — everyone works the module on the virtual arm, up to the step that needs hardware.
• Real lane — one pair at a time takes the physical arm for that step.

Keep the rotation to a visible timer, and make the waiting lane productive: the reading, the theory and the simulator tasks are genuinely the same work.

If the calibration has been exported and imported in advance, a changeover costs seconds instead of a recalibration.

What goes wrong, by module

Module 1 — motor IDs. They connect all six motors and then try to assign IDs. The app tells them, but the message arrives after the frustration. Say it first: one motor on the bus at a time.

Module 3 — the HOME pose. They eyeball it, press Calibrate, and inherit an offset for the rest of the course. Check poses before anyone presses the button.

Module 4 — the enable key. They forget to hold left shift and conclude the arm is broken.

Module 5 — grip force. They crush the object, then over-correct and drop it. The force scale is on screen; point at it.

Module 6 — the four-point tool calibration. Fiddly. The arm has to touch the same physical point from four genuinely different orientations, and learners tend to move it barely at all. Demonstrate it.

Module 7 — HSV tuning. They drag the sliders at random. Give them the order: hue first to find the colour band, then saturation, then value.

Module 8 — PID. They set a huge gain, watch it oscillate, and are delighted. That is the lesson; let it happen, then ask them to damp it.

Module 9 — the dataset. They shoot sixty near-identical frames and the lesson rejects them. This is the intended failure. Ask what would make a robust dataset before they reshoot.

Housekeeping that saves the next session

• Park each arm in a low, folded pose before powering off.
• Coil the USB cables with the arms, not in a shared box: a charge-only cable that finds its way into the pile costs someone twenty minutes.
• Keep one spare calibrated arm entry exported to a file; a rebuilt arm is back in service in a minute.
• Note which pairs finished. It is the only progress record that exists on your side of the platform.