Alarms
An operator watching twenty robots can act on only a few things at once. Alarms decide which, and how loudly. The rules follow ISA-18.2 and IEC 62682, the standards for industrial alarm management.
Which surface
Ask two questions: must someone act, and is it about the robot or about this page?
| Surface | Use when | Lasts | Component |
|---|---|---|---|
| Emergency stop | The robot has been stopped, or must be. | Until released with Slide to release. | stop-button, island |
| Alarm | An abnormal condition an operator must act on. | Until it clears and is acknowledged. | alarm |
| Alert | A condition on this page, true for a while. | While the condition holds. No acknowledge. | alert |
| Toast / island notice | The result of what the user just did, or a minor event. | 4 s · 8 s with Undo · failures until dismissed. | toast, island |
- Emergency stop
- Use when
- The robot has been stopped, or must be.
- Lasts
- Until released with Slide to release.
- Component
stop-button, island
- Alarm
- Use when
- An abnormal condition an operator must act on.
- Lasts
- Until it clears and is acknowledged.
- Component
alarm
- Alert
- Use when
- A condition on this page, true for a while.
- Lasts
- While the condition holds. No acknowledge.
- Component
alert
- Toast / island notice
- Use when
- The result of what the user just did, or a minor event.
- Lasts
- 4 s · 8 s with Undo · failures until dismissed.
- Component
toast, island
Three priorities
Each has its own shape, colour and word, so priority reads without colour vision. Solid red stays reserved for the emergency stop.
| Priority | Act | Sound and haptics | Share of alarms | Examples |
|---|---|---|---|---|
| Critical | Now | Every 10 s until acknowledged | ≤ 5 % | Collision risk, person in a restricted zone, robot fell |
| High | Within minutes | Plays once | ≈ 15 % | Battery below 15 %, localisation lost, stuck for 30 s |
| Low | When convenient | Silent | ≈ 80 % | Cleaning due, firmware update ready, map out of date |
- Critical
- Act
- Now
- Sound and haptics
- Every 10 s until acknowledged
- Share of alarms
- ≤ 5 %
- Examples
- Collision risk, person in a restricted zone, robot fell
- High
- Act
- Within minutes
- Sound and haptics
- Plays once
- Share of alarms
- ≈ 15 %
- Examples
- Battery below 15 %, localisation lost, stuck for 30 s
- Low
- Act
- When convenient
- Sound and haptics
- Silent
- Share of alarms
- ≈ 80 %
- Examples
- Cleaning due, firmware update ready, map out of date
The shares are the ISA-18.2 guide. If a third of your alarms are critical, the priorities have stopped meaning anything: review them.
Lifecycle
Acknowledging and clearing are different events. An alarm leaves the list only when both have happened.
High: Localisation degraded · Floor 3
1 · Raised. The condition starts. Strongest form; sound plays for critical and high.
| State | Who sets it | In the list |
|---|---|---|
| Shelved | An operator, for a set time (not critical alarms) | Out of the live list into “Shelved”, silent, with Unshelve. Returns when the time is up. |
| Suppressed by design | The app's logic: the condition does not apply now, e.g. a docked robot is not moving | Counted under the list |
| Out of service | An authorised person, for maintenance | Counted under the list |
| Flood | The list: over 10 alarms per operator raised in 10 minutes (ISA-18.2). A design fault to fix at the source, not a load to carry | Named in a banner; critical and high stay in view, low alarms fold into one row; screen readers hear one count |
| Not known | The app: the alarms are still loading, or could not be loaded or updated | Placeholder rows while loading, never “No active alarms”. On a failure, a danger-tint banner says the list may be out of date (or that alarms may be missing), when it last updated, and offers Try again; alarms already loaded stay |
- Shelved
- Who sets it
- An operator, for a set time (not critical alarms)
- In the list
- Out of the live list into “Shelved”, silent, with Unshelve. Returns when the time is up.
- Suppressed by design
- Who sets it
- The app's logic: the condition does not apply now, e.g. a docked robot is not moving
- In the list
- Counted under the list
- Out of service
- Who sets it
- An authorised person, for maintenance
- In the list
- Counted under the list
- Flood
- Who sets it
- The list: over 10 alarms per operator raised in 10 minutes (ISA-18.2). A design fault to fix at the source, not a load to carry
- In the list
- Named in a banner; critical and high stay in view, low alarms fold into one row; screen readers hear one count
- Not known
- Who sets it
- The app: the alarms are still loading, or could not be loaded or updated
- In the list
- Placeholder rows while loading, never “No active alarms”. On a failure, a danger-tint banner says the list may be out of date (or that alarms may be missing), when it last updated, and offers Try again; alarms already loaded stay
Touch and sound
Gloves, wind and machine noise: the screen alone cannot tell an operator that a stop went out or a critical alarm came in. Each safety moment has one fixed pattern from @syncrobotic/, so the hand learns it like the eye learns a colour. Both the phone's island and the dashboard's list signal; a critical alarm sounds again every 10 s until someone acknowledges it, and shelved, suppressed and out-of-service alarms stay silent. Touch works where the browser can vibrate (Android, not iOS); sound only after the first tap on the page. Neither ever replaces what the screen shows.
| Moment | Vibration (ms on, off, on…) | Tone |
|---|---|---|
stopSent | 40 | None |
stopConfirmed | 40, 80, 40 | None |
stopUnconfirmed | 200, 100, 200 | None |
release | 30 | None |
driveStart | 15 | None |
driveStop | 15, 60, 15 | None |
alarmCritical | 200, 100, 200, 100, 200 | 5 pulses of 100 ms at 880 Hz |
alarmHigh | 120, 80, 120 | 3 pulses of 120 ms at 660 Hz |
stopSent- Vibration (ms on, off, on…)
- 40
- Tone
- None
stopConfirmed- Vibration (ms on, off, on…)
- 40, 80, 40
- Tone
- None
stopUnconfirmed- Vibration (ms on, off, on…)
- 200, 100, 200
- Tone
- None
release- Vibration (ms on, off, on…)
- 30
- Tone
- None
driveStart- Vibration (ms on, off, on…)
- 15
- Tone
- None
driveStop- Vibration (ms on, off, on…)
- 15, 60, 15
- Tone
- None
alarmCritical- Vibration (ms on, off, on…)
- 200, 100, 200, 100, 200
- Tone
- 5 pulses of 100 ms at 880 Hz
alarmHigh- Vibration (ms on, off, on…)
- 120, 80, 120
- Tone
- 3 pulses of 120 ms at 660 Hz
Try these on an Android phone: iOS Safari does not vibrate, and sound plays only for the two alarms.
Rules
- No action, no alarm. If the operator cannot do anything about it, log it as an event.
- One alarm per root cause. When a robot goes offline, raise “offline”, not also “camera lost” and “lidar lost”.
- The title says what happened and where (“Collision risk · Hall B”); the source names the robot; the action says what to do (“Open camera”).
- Never auto-dismiss an alarm and never send one as a toast.
- No blinking and no tinted rows. Colour is only on the glyph and the priority word. Unacknowledged is a bold title and a solid bar; cleared but not acknowledged keeps the bold with a dashed bar; acknowledged is regular and grey and says when. Sound does the interrupting.
- Time first, in tabular figures, so order and age read down the column.
- “Acknowledge all” is the app's choice (onAcknowledgeAll) and never takes a critical alarm: each critical alarm is read and acknowledged on its own row. It takes only the other alarms in view, never shelved, suppressed or out-of-service ones, and says how many it acknowledged.
- New critical alarms are announced to screen readers assertively, others politely.
AlarmListdoes this for you. - In the control app the island shows the most urgent unacknowledged alarm (
IslandAlarm), above any notice, and its edge takes the alarm's colour. The dashboard shows the fullAlarmList. Both read the same data.
Critical: Collision risk · Hall B
Error
Toast
Feedback on what the user just did. Same timings as the island's notices in the control app.