Bad instability
Advice changes for the wrong reasons
Wrong data
Driver cannot do what is requested
Missing integrations
Unknown parameters stay outside the model
Slow calculation
Advice arrives obsolete
Knowledge base
Why stable driver advice still matters
Stable advice is straightforward in principle: avoid bad recalculations, react fast to real railway changes, and keep enough driver trust for the system to matter operationally.
Published: January 10, 2026
TopicsAdvice stabilityHuman factorsDriver trustOperational reliabilityDAS
Why advice changes
Some changes are bad, some are valid
Wrong data, missing parameters, and slow calculation are bad reasons. Timetable, route, temporary constraints, and signalling changes are valid reasons.
Valid recalculation
Advice changes because the railway changed
New timetable
New route
New temporary constraints
New signalling data
Meaning
The goal is not to freeze advice. The goal is to prevent unnecessary change and react immediately when the real operating context changes.
Prevention
What keeps advice believable
This is mostly engineering work: better data, faster loops, better localisation, and better physical modelling.
Prevention
Validate data
Check source data against real operation so the driver is not asked to do impossible things.
Prevention
Direct driver reporting
Let drivers report bad advice quickly and push that feedback back into the source systems.
Prevention
Fast onboard calculation
Keep latency low enough that advice is still current when it reaches the cab.
Prevention
Sensor fusion localisation
Improve localisation so timing and route context stay precise.
Prevention
Accurate train performance models
Use train-specific behaviour instead of generic rolling-stock assumptions.
Prevention
Detailed physical inputs
Model more than gradients and speed limits: power restrictions, low adhesion, powermaps, curves, and tunnel factors.
Operational consequence
When trust drops, the network-level premise drops with it
If drivers stop using the advice, too few trains stay inside their allocated paths and the whole premise weakens.
Step 1
Unstable or implausible advice
Advice changes too often, arrives too late, or simply does not make sense to the driver.
Next
Step 2
Trust drops
Drivers stop believing the system is worth following when it asks for things that do not fit the real situation.
Next
Step 3
Drivers stop using it
The advisory may still calculate, but operationally it no longer has enough real usage to shape behaviour.
Next
Step 4
Not enough trains stay inside allocated paths
Without critical mass, the network does not get the path discipline the advisory assumes.
Next
Step 5
Conflicts remain and advice can become invalid again
The premise weakens because trains still interfere with each other and downstream advice can quickly lose validity.
System implication
Stability matters because advisory only works if enough trains actually follow it. If trust collapses, the network-level effect collapses with it.
Brass tacks
- Bad instability comes from wrong data, missing integrations that hide needed parameters, or calculation that is too slow to stay current.
- Good advice comes from data validation, direct driver reporting, low-latency onboard calculation, exact localisation, accurate train models, and detailed physical inputs.
- If drivers stop trusting the advice, too few trains stay inside their allocated paths, conflicts remain, and downstream advice can lose validity again.
Source material
Related articles