Observed, derived, assumed, and fallback inputs

Road Risk separates retrieved data, derived calculations, user-selected settings, and fallback assumptions so the model can be inspected.

Open App V2 Validation boundary

How to scan this register

Read the summary before the table

Page Purpose

How to scan this register

This page explains what is observed, calculated, selected, or substituted before the full assumption table appears.

Observed inputs: Public road geometry and tags where available.
Derived values: Radius, curvature, stopping distance, safe-speed checks, and local bin context.
Selected settings: Weather, lighting, surface, vehicle, behaviour, and speed assumptions chosen for scenario testing.
Fallbacks: Visible substitutions used when public data is incomplete; these should make interpretation more cautious.

Assumption groups

Assumptions and their interpretation role

Vehicle

Vehicle profile assumptions affect stopping and handling interpretation. They are adjustable but not a full vehicle simulation.

Surface and friction

Surface tags and friction assumptions affect safe speed, slip-style indicators, and braking distance.

Weather and visibility

Rain, fog, lighting, and visibility assumptions change friction and perception context; they are not live weather measurements.

Behaviour

Fatigue, distraction, BAC, and overspeed are controlled scenario settings, not claims about actual drivers.

Infrastructure and context

Road class, lanes, lighting, pedestrian/cycle context, medians, and barriers depend on public tag completeness.

Fallback and confidence

Fallback values keep the model inspectable when data is missing, but should lower interpretation confidence.

Most important Phase 1 assumption records

Assumption	Affects	Input type	Limitation
Vehicle profile	Stopping distance, handling interpretation, visibility/context assumptions.	Assumed / user-selected	Not a full vehicle simulation.
Friction coefficient / surface	Safe speed, slip ratio, braking distance.	Observed if tagged; fallback otherwise	Actual tyre-road friction is not measured live.
Reaction time	Reaction distance and total stopping distance.	Assumed	Does not observe individual driver response.
Rain / wet road	Friction, stopping distance, model multiplier.	Scenario assumption	Not live rainfall intensity.
Fog / visibility	Visibility context and reaction interpretation.	Scenario assumption	Does not measure actual sight obstruction.
Lighting	Visibility, infrastructure context, confidence notes.	Observed if OSM lit tag exists; fallback otherwise	Missing tag does not prove absence of lighting.
Fatigue / distraction	Behavioural multiplier and reaction interpretation.	Scenario assumption	Not a measurement of real driver state.
BAC / alcohol impairment	Behavioural multiplier and stopping-distance interpretation.	Scenario assumption	Does not infer impairment at any location.
Overspeed	Centripetal demand and velocity-squared stopping distance.	Scenario assumption	Not evidence of actual speeding.
Road classification	Traffic proxy, speed context, baseline exposure assumptions.	Observed OSM highway tag; fallback if unavailable	Road class is an imperfect exposure proxy.

Questions judges may ask

Most important assumptions to explain

Friction / surface grip

Lower friction reduces safe speed and increases braking distance. It is one of the clearest sensitivity levers.

Reaction time

Reaction time is assumed, not observed, so fatigue and distraction should be treated as controlled scenarios.

Road class

OSM road class helps provide context but is not a measured traffic-exposure dataset.

Visibility / lighting

Visibility assumptions affect interpretation, but do not prove actual visibility at a road.

Missing tags

Missing public map tags do not prove a feature is absent; they show where evidence quality is weaker.

Sensitivity guidance

How to interpret assumption changes

Fallback and confidence

Fallback values keep the model inspectable when public data is missing, but they should lower confidence in the interpretation and remain visible in saved records.

Controlled comparison

Assumption sensitivity is clearest when the same road is compared with one setting changed at a time. It does not prove that the selected condition actually occurred.

Change one setting at a time when explaining model behaviour.
Treat fallback-heavy cases as lower-confidence records.
Read weather and behaviour settings as selected assumptions, not claims about actual conditions.
Compare scenarios only when the same road and model settings are otherwise held stable.