Home / Platform / ADAS Configuration
Driver assistance, on your terms

ADAS is only as useful as the thresholds behind it. Those are yours to set — and to show.

Forward-collision, lane-departure, door-zone and driver-state detection, configured per vehicle class and route from the platform, calibrated on the line, and deployed to the camera or edge recorder over the air — with every event arriving on the same feed as the rest of the vehicle.

Which page is which. In-Vehicle ITS names the video AI functions that run on the vehicle. This page covers how those functions are configured, calibrated, deployed and held to account.
Functions

Four detections, and nothing implied beyond them.

Forward collision

Closing-distance events from the forward-facing camera.

Lane departure

Unsignalled lane exit events.

Door-zone safety

Detection in the door zone during dwell.

Driver monitoring

Drowsiness, distraction and phone use, from the driver-facing camera.

Configuring the events

A 12 m city bus and a 3 W feeder do not deserve the same sensitivity.

Configuration is the whole product here. A default threshold applied to a whole fleet produces either noise nobody reads or silence that misses the event.

Thresholds per vehicle class and route

Sensitivity and severity set for the vehicle it is running on. Profiles are held on the platform and applied by model variant, route and time band.

Geofenced behaviour

Depot yards, terminals and known slow corridors carry different sensitivity from arterial running — so the yard does not generate a day's worth of forward-collision noise before the first trip.

Severity and escalation

Each event class carries a severity, a driver-facing alert — audio via the PA, visual via the driver console where one is fitted — and a control-room escalation rule.

Explainability

Every threshold is visible and adjustable, and an alert can always be traced to the limit that fired it. The same rule-based posture the Intelligence page describes, applied to vision events.

Calibration and deployment

From the line to the device.

Line calibration

Camera position, height and field of view calibrated as part of end-of-line verification, with the calibration record bound to the VIN. The same station described on OBU to Cloud.

Signed profiles

Configuration is deployed as signed, version-controlled profiles through the configuration engine, with rollback.

Over the air

Where the vehicle runs the RouteSync OBU software stack, configuration and model updates travel over the same FOTA channel as everything else — staged and reversible.

Model source

The vision models can be the camera vendor's or RouteSync's. Where they are ours, model updates travel over FOTA with the configuration. Where the camera or recorder is a vendor unit running the vendor's models, deployment goes through the vendor's documented interface, per the scope boundary on OBU to Cloud.

Events on the platform

What happens after the camera raises an event.

A detection that stops at the device is a beep in a cab. These are the five things that make it part of the operation.

Ingested like any other event

Through OBU to Cloud — authenticated, normalised, reconciled and stored on the same path as position and telemetry.

Into driver behaviour scoring

As a weighted, visible input. The weight is set by you and can be taken apart, which is the difference between a score a driver can argue with and one they cannot.

Evidence clip retrieval

Configured event classes trigger clip retrieval from the MDVR, so the footage exists before anyone asks for it.

Into the incident queue

Events enter the control-room console with acknowledgement and closure audit, alongside every other exception.

Onto the boards

They appear on management boards under the same green, amber and red logic as every other KPI — from configured limits, not a separate scale.

Scope

What ADAS here does not do.

What ADAS here does not do. It warns; it does not intervene. No function on this page brakes, steers or overrides the driver — that is a vehicle-level system homologated with the vehicle, not something a retrofitted camera should claim. Detection quality depends on the model, the mounting and the light, so we calibrate and validate it per vehicle and per camera position rather than quote a figure. Where the camera or edge recorder is a closed third-party unit, configuration and model deployment go through whatever interface the vendor exposes, and functions that interface does not reach are out of scope until it does.

Bring a route that generates too many alerts.

Thresholds are the argument worth having, and they are easier to have against a corridor your team already complains about.

Thanks — we'll be in touch shortly.