When the head gets stuck: two causes, four safeguards
- Log
- 21 of 25
- Date
- Reading
- 2 min read
Skills in this log
Something broke in this log. It is written up anyway.
Thursday 5 November, early afternoon: the robot had its head down, resting against its body. It heard things, it tried to turn, nothing moved. Late afternoon, same again. Going back through the log, I found two more episodes, on 29 October and 2 November.
What happened
The robot has two resting postures where the head comes down almost to the body: waiting and sleeping. Two things combined on waking up.
A gesture that copied the wrong starting point. When a sound startles it, it glances towards the sound, then comes back. To come back, it read the position of its head. Right after sleep, that position was a head tilted by about 60°. The gesture copied it and added 25 to 28° of rotation. So the head set off towards a pose the mechanics cannot hold.
A start that was too abrupt. The maker knows the case: a public ticket describes the lock-up when a movement starts suddenly from one of these resting postures. It suggests a simple remedy: give the head one second to reach the start of the movement.
Freeing the head, if it happens to you
Switch the motors off, lift the head by hand to the upright position, switch the motors back on, then play the wake-up. That is what worked here, and it is the same manoeuvre the ticket describes. Do not force the head with the motors on.
The four safeguards
- Tilt is capped in every gesture: 20° at most, and 10° when the head turns by more than 20°.
- Waking up re-centres the head before any other gesture, and the log reports a head left down.
- A watchdog compares the target with the position reached. Two gaps of more than 10° in a row, and gestures stop for five minutes.
- A soft transition: if a movement starts far from its first frame, the head first reaches it in one second. Close to that frame, nothing changes, and the startle keeps its speed.
What I take from it
In both causes, the robot did exactly what it was asked. The first was a mistake in my program. The second is known to the maker, and it was still my program that triggered it. After the update, the head straightened up in a few seconds at start-up, without a single error.
Next
Once the head was free, something else came back: a slight tremor of the body, which I thought was fixed. I will come back to it.
