Expected states
A task reported as done does not prove that its effect is there. This strategy reads what the function of the task expects (“Extinguish a barbecue” leaves the barbecue Off) and compares it with the state the sensors give the instance on the field. The task is done, the barbecue is still On: that is a Failing deviation.
The state asked, the state observed
On the task — the “From Instance” rows of its consumes table that carry an expectedState. A “From Concept” row of the function becomes such a row at deduction, on the instance the problem impacts: see what a function asks.
On the field — the current state of the ACTIVE instance, or its default state when it has none yet. Only the server writes this state, from the sensors, through the state machine of the asset.
One contribution per instance out of state
for each From Instance row with an expected state:
if the instance is on the field and its state ≠ expected → contribution mismatch / threshold on Failing
The grace — the field is given some time after the end of the task: a barbecue takes a while to cool down. Before the grace is over, the task is not judged.
The message — each contribution is on the field world and names the task, the instance and both states: "Extinguish a barbecue" is done but "Martinez barbecue" is On, not Off. With nothing to report, Failing receives 0 from this strategy.
Default configuration
Defaults of getDefaultConfiguration. No configuration form ships in supervision-strategy-extensions.json for this strategy: the supervision box only checks or unchecks it.
| Parameter | Key | Default | Meaning |
|---|---|---|---|
| Grace | grace | 30 | Seconds given to the field after the end of the task before its state counts. |
| Mismatch | mismatch | 1.0 | Weight of an instance out of its expected state; at 0 or below, nothing is counted. |
| Threshold | threshold | 1.0 | Weighted sum at which Failing reaches 1. |
What it does not judge
States only — the expected sensor value (expectedValue) and the expected position (expectedPosition) of a consumes row are read by the knowledge consolidation and by the simulation, not by this strategy.
An instance without state — with neither a current nor a default state, or absent from the field, the row is skipped: no deviation can be established.