State Reconstruction¶

State reconstruction is a disciplined observer hypothesis, not machine truth.

A Concrete Artifact¶

# synthetic passive observer workbook
t=12:00:00.000  DBC ENGINE_STATUS.EngineSpeed = 1000 rpm  source=samples/first-workbook/expected/dbc.yml
t=11:59:57.000  DBC ENGINE_STATUS.EngineState = Off       source=samples/first-workbook/expected/dbc.yml
t=12:00:00.500  DBC ENGINE_STATUS.EngineLoad = 45 percent source=synthetic future-dated record
expected freshness window: 1 s
expected signal: HydraulicPressure
t=11:59:59.900  EngineSpeed = 850 rpm from synthetic alternate authority

Synthetic offline sample for explanation; not a real operational trace or live-system instruction.

Inspection Trap¶

A decoded value can look like state even when it is only a recent observation under one authority. The trap is to collapse freshness, missingness, source confidence, and conflict into a single convenient truth value.

This page inspects synthetic workbook evidence as observer state: one signal can be fresh, stale, missing, or unknown, and it can also be conflicting or low-confidence depending on timestamp, provenance, and expected evidence.

Worked Decode¶

1. The DBC fixture in samples/first-workbook/ maps CAN identifier 0x123 to ENGINE_STATUS.
2. The matching synthetic payload decodes EngineSpeed as 1000 rpm and EngineState as Off.
3. At observer time 12:00:00, EngineSpeed was observed 100 ms ago, so it is fresh under a 1 s freshness window.
4. EngineState was observed 3 s ago, so it is stale under the same window.
5. HydraulicPressure is expected by the workbook DBC but absent from this observer state, so the assessment is missing.
6. EngineLoad carries a timestamp later than the observer time. Its evidence cannot be ordered on this time base, so its assessment is unknown with zero confidence.
7. A near-contemporaneous second EngineSpeed observation of 850 rpm, with the same unit but a different synthetic authority, creates a conflict under the stated overlap rule. The conflict does not prove which value is right.

What The Evidence Supports¶

• The artifact supports explicit observer assessments for fresh, stale, missing, conflicting, and low-confidence evidence.
• It supports source-aware reasoning over decoded synthetic values, timestamps, expected signals, and alternate authority.
• It supports a cautious state hypothesis with reasons, not a hidden-machine-state claim.

What The Evidence Does Not Support¶

• The artifact does not prove physical state, sensor health, safety effect, actuator intent, or application consumption.
• It does not choose the correct value when sources conflict.
• It does not replace domain-specific timing, redundancy, synchronization, or certification analysis.

Passive Reconstruction Cases¶

Field Layout / Anatomy¶

Visual Model¶

Timing And Authority¶

Timing authority comes from the observer clock and the expected freshness window. Semantic authority comes from the DBC or another description source. State authority is weaker than both: it is a hypothesis over observed evidence. A fresh decode can still be semantically wrong if the authority is wrong, and a semantically correct decode can still be stale. Evidence dated after the observer time is not extra-fresh evidence; it is an ordering anomaly that remains unknown until the time bases are reconciled.

Bytes rarely explain themselves. Name the source that defines meaning before naming engineering values: standard metadata, label table, DBC, PGN/SPN definition, object dictionary, register map, DID catalog, LDF, ARXML, channel metadata, or vendor profile.

Failure And Ambiguity¶

• Missing can mean never observed, filtered out, outside the capture window, or not scheduled in the current mode.
• Stale can mean the machine stopped publishing, the capture missed traffic, a gateway cached a value, or the expected freshness window is wrong.
• A future timestamp can indicate clock disagreement, metadata conversion, or evidence from another time base. The observer does not infer which explanation is correct.
• Conflict requires comparable units, distinct sources, timestamp evidence, and a stated overlap window. Same-source evolution is not conflict. Remaining disagreements can still come from wrong authority, clock error, gateway conversion, or a real state transition seen through independent sources.
• Confidence decay is observer confidence only. It does not prove sensor health, safety effect, physical truth, actuator intent, or application consumption.

Python Model¶

The current package exposes a concrete passive reconstruction example:

"""Runnable passive state reconstruction example for the schoolbus binder."""

from datetime import UTC, datetime, timedelta

from schoolbus.core import Observation, Signal, Unit
from schoolbus.formats.dbc import DbcDatabase
from schoolbus.observe import PassiveObserver, assess_latest, find_conflicts
from schoolbus.protocols.examples import DBC_TEACHING_TEXT

now = datetime(2026, 5, 21, 12, 0, tzinfo=UTC)
max_age = timedelta(seconds=1)

database = DbcDatabase.from_text(
    DBC_TEACHING_TEXT,
    source="samples/first-workbook/raw/dbc_engine_status.dbc",
)
signals = database.decode(0x123, bytes.fromhex("00 40 1f 40 1f 00 00 00"))

observer = PassiveObserver()
observer.observe(
    Observation(
        signal=signals["EngineSpeed"],
        source="samples/first-workbook/expected/dbc.yml",
        timestamp=now - timedelta(milliseconds=100),
    )
)
observer.observe(
    Observation(
        signal=signals["EngineState"],
        source="samples/first-workbook/expected/dbc.yml",
        timestamp=now - timedelta(seconds=3),
    )
)

state = observer.state()
print(assess_latest(state, "EngineSpeed", max_age, now=now).show_fields())
print(assess_latest(state, "EngineState", max_age, now=now).show_fields())
print(assess_latest(state, "HydraulicPressure", max_age, now=now).explain())

conflicts = find_conflicts(
    [
        Observation(
            signal=signals["EngineSpeed"],
            source="samples/first-workbook/expected/dbc.yml",
            timestamp=now - timedelta(milliseconds=100),
        ),
        Observation(
            signal=Signal(
                "EngineSpeed",
                850,
                Unit("rpm"),
                source="synthetic alternate authority",
            ),
            source="synthetic alternate authority",
            timestamp=now - timedelta(milliseconds=100),
        ),
    ]
)
for conflict in conflicts:
    print(conflict.show_fields())
    print(conflict.explain())

The model keeps assessment local and explicit: one signal, one status, one confidence score, one reason. It reports conflicts without selecting a winning source.

See the passive state reconstruction lab for the runnable workbook view.

Source Notes¶

Source confidence: High for project-local observer concepts, synthetic workbook evidence, and API behavior. Physical state and signal semantics remain authority-dependent.

The confidence score is deliberately linear and local to this workbook. Real systems may require richer timing, source-quality, synchronization, redundancy, and domain models before making operational claims.

References¶

Repo-Owned Context¶

• docs/governance/project-charter.md
• docs/concepts/observer-model.md
• docs/binder/foundations/observer-reconstruction.md
• docs/binder/foundations/timing-and-cadence.md
• samples/first-workbook/

Public Cross-Checks¶

This page is project pedagogy over synthetic evidence. Protocol-specific public cross-checks live on the relevant binder pages, especially CAN and DBC.

Project posture is aggregated in the protocol support policy, source policy, and project charter.