How MRT Communications and Passenger Information Work Using Mathematics

A passenger-information message can be factually correct and still fail the passenger.

If the train state changed thirty seconds ago and the screen still describes the old state, correctness has expired.

An MRT moves people through a physical network.

Passenger information moves decisions through an information network.

The passenger sees:

the next station on an in-train screen,
which side the doors will open,
a train-arrival estimate,
a service-disruption notice,
an alternative route,
a station poster, app notification or staff instruction.

Behind those messages is a state pipeline:

railway state
→ observation
→ interpretation
→ message generation
→ communication channel
→ display / announcement / staff
→ passenger receives
→ passenger decides
→ passenger movement changes

Singapore’s MRT already uses multiple layers. MOT has stated that MRT train models provide route maps, direction of travel, current and upcoming stations and the opening side of doors through LCD screens, in-train announcements and other displays. LTA and operators also publish train-status and disruption information through MyTransport.SG, websites, social channels and station staff.

In January 2026, LTA and the rail operators explicitly strengthened commuter-centric disruption communications and alternative-travel guidance as part of Rail Reliability Taskforce service-recovery measures.

This article owns that information layer. It does not own safety-critical signalling communications or movement authority; those belong to How MRT Signalling and Train Regulation Work Using Mathematics. The permanent whole-system hub is How MRT Works | It’s Mathematics.

The RFE — What Is Passenger Information For?

The weak objective is:

display as much information as possible

Too much information can increase confusion.

The opposite failure is too little information.

A disruption without useful guidance turns passengers into independent guessers, often concentrating them onto the same obvious alternative.

The Reason for Existence is:

give the passenger the smallest timely message that changes the next decision correctly enough to preserve the journey.

Prompt 1 — What Makes Information Fresh?

Suppose a railway state was measured at time t_m.

The passenger sees the message at time t_r.

information age τ = tr − tm

A message can be perfectly accurate at t_m and stale at t_r.

For a changing quantity x(t), stale-state error is approximately:

estale ≈ x(tr) − x(tm)

If the state changes roughly at rate dx/dt:

estale ≈ (dx/dt)τ

The faster the railway state changes, the more expensive latency becomes.

A five-minute-old planned closure notice may still be useful.

A five-minute-old train-arrival estimate can be useless during a disruption.

Freshness is not a universal number. It depends on how quickly the decision-relevant state can change.

Prompt 2 — How Is an Arrival Time Predicted?

A simple arrival estimate is remaining distance divided by speed.

ETA ≈ remaining distance / average future speed

But a real train may still need to:

complete the current dwell,
wait for movement authority,
accelerate,
coast,
brake,
respond to the train ahead,
serve another station.

A more useful prediction is:

Tarrival
= tnow
+ predicted dwell remaining
+ predicted run time
+ predicted control delay

The prediction should also carry uncertainty.

ETA = 2.0 min ± uncertainty

Public displays often compress uncertainty into rounded minutes rather than showing a confidence interval.

That compression is useful only if the underlying estimate is calibrated.

Prediction error is:

eETA = Tactual − Tpredicted

Across many arrivals:

MAE = (1/n) Σ |eETA|

A good information system should learn whether its estimates are systematically early, late or unstable by station and operating state.

Prompt 3 — Why Must Every Channel Tell the Same Story?

During a disruption, passengers may receive information from:

platform screens,
public-address announcements,
in-train displays,
MyTransport.SG,
operator websites,
social media,
station staff.

Let message state on channel c be M_c(t).

A simple consistency condition is:

Mc(t) ≈ M*(t)

where M* is the current authoritative public message.

Cross-channel inconsistency can create a decision split:

screen says use Line A
app says Line A disrupted
staff says use shuttle bus
→ passenger confidence collapses

The system therefore needs message versioning.

message = [incident ID, version, timestamp, affected area, advice]

The passenger should not need to reconcile conflicting versions manually.

Prompt 4 — How Does Information Change Passenger Flow?

Information is not passive.

It changes route choice.

Suppose two alternatives have perceived costs C_A and C_B.

A simple logit route-choice model is:

P(A) = exp(−θCA) / [exp(−θCA)+exp(−θCB)]

If a message says Route A now requires 20 extra minutes, perceived C_A rises and more passengers choose B.

But that changes B’s crowding.

information
→ passenger rerouting
→ alternative-route load rises
→ waiting and crowding change
→ best advice may change again

This creates an information-feedback problem.

LTA’s 2026 disruption-recovery measures explicitly emphasise clearer alternative-travel guidance and commuter-centric communications. The point is not merely to tell passengers that something failed; it is to distribute human demand across the remaining transport network intelligently enough to support recovery.

A disruption message is a control input to the passenger network.

Prompt 5 — How Much Information Is Too Much?

More data is not always more useful.

A passenger deciding whether to remain on a platform may need only:

what is affected
how long the extra journey may be
what alternative is available
where to go now

They do not need a full engineering incident report.

Message usefulness can be represented:

Utility(message)
= decision improvement
− cognitive load
− ambiguity
− latency cost

Good communication therefore compresses.

It should not omit a condition that changes the decision, but it should omit detail that does not.

Prompt 6 — What Happens When the Information System Fails?

Failure modes include:

no message,
stale message,
wrong affected area,
conflicting channels,
correct incident but poor alternative advice,
message inaccessible to some passengers,
information arriving after the passenger has already committed to a route.

Define decision loss L:

L = journey cost under received information
  − journey cost under best available information

Passenger-information reliability is therefore not merely uptime.

A screen can have 100% electrical uptime while continuously showing stale information.

We need at least:

availability
× correctness
× freshness
× consistency
× accessibility

as the public information function.

Prompt 7 — Why Does Accessibility Apply to Information Too?

Information can be inaccessible even when the station is physically accessible.

Passengers differ in:

vision,
hearing,
language,
mobility,
familiarity with the network,
digital access.

A visual-only disruption notice is weak for a passenger who cannot see it.

An audio-only message is weak for a passenger who cannot hear it.

An app-only solution excludes passengers without a usable device or connection.

Information redundancy should therefore be multimodal.

visual + audio + physical wayfinding + staff + digital

This is functional redundancy, not duplication for its own sake.

The same message may need different representations for different receivers while preserving the same underlying claim.

Prompt 8 — How Does Information Recover After a Disruption?

A disruption message has a lifecycle.

incident detected
→ affected service confirmed
→ initial public message
→ alternative route guidance
→ updates as state changes
→ partial restoration
→ service restored
→ residual passenger recovery
→ message closed

The last step matters.

If normal service is restored but old warnings remain visible, the information system can continue routing passengers away from a route that is already usable.

Recovery therefore requires information withdrawal as well as information publication.

A message should have validity:

valid from t0
until state changes or explicit expiry

The information system itself needs World Return:

message predicted to reduce confusion
→ passengers move
→ observe crowding / questions / route choices
→ revise message if required

A Complete Fictional Disruption-Information Example

A fictional line loses service across three stations.

There are 6,000 passengers approaching the affected area over the next 30 minutes.

Alternative A can absorb 2,500 additional passengers.

Alternative B can absorb 2,000.

Shuttle buses can absorb 1,000.

alternative spare capacity
= 2,500 + 2,000 + 1,000
= 5,500 passengers

The remaining 500 passengers cannot all be absorbed immediately.

If every public message recommends Alternative A, that route overloads while B remains underused.

A better information policy distributes passengers:

origin group 1 → A
origin group 2 → B
short local trips → shuttle
others → delay journey where feasible

Suppose initial message latency is four minutes and passenger arrival rate is 200/min.

Before guidance reaches passengers:

unredirected passengers ≈ 4×200 = 800

Latency has become crowding.

Reducing message latency from four minutes to one minute prevents roughly 600 passengers in this fictional case from entering the obsolete decision path.

The Communications Deletion Tests

Remove timestamps: passengers cannot distinguish current from stale information.
Remove source authority: conflicting messages have no canonical owner.
Remove latency: the model pretends information arrives instantly.
Remove uncertainty: arrival estimates look exact when conditions are unstable.
Remove channel consistency: app, screen and staff can direct passengers differently.
Remove accessibility: one message format is treated as usable by everyone.
Remove passenger feedback: route advice never notices that it has overloaded the alternative.
Remove expiry: obsolete disruption guidance survives after recovery.

The Information Paradoxes

A correct message can be wrong for the passenger if it arrives too late.
More information can reduce understanding.
The best alternative route can stop being best after everyone is told to use it.
A screen can be available while the information service is functionally failed.
Removing an old warning can be as important as publishing a new one.
Different passengers may need different representations of the same underlying state.

The Passenger-Information Audit

What passenger decision must this message support?
What railway state is the message based on?
When was that state measured?
How old will the information be when received?
How quickly can the relevant state change?
What uncertainty belongs to the prediction?
Which source owns the authoritative public message?
Are all channels on the same version?
Can visual, audio, staff and digital channels reach different receivers?
What alternative route capacity exists?
Will the advice overload the route it recommends?
What message can be removed without changing the decision?
When should the message expire?
How will passenger behaviour show whether the advice worked?

The World Return — When Passengers Answer the Message

The information system predicts:

ETA
alternative route load
passenger understanding
queue reduction

Reality returns:

actual arrival
actual passenger redistribution
actual station questions
actual queues

If arrival predictions are systematically optimistic, recalibrate.

If an alternative route becomes overcrowded after advice is issued, update the advice.

If passengers repeatedly ask staff the same question despite a screen message, the message has not succeeded merely because it was technically displayed.

Passenger information works when the passenger’s next action becomes better—not when the system can prove that it sent a message.

Conclusion — Information Is Part of the Railway’s Control Surface

A train changes state.

The information system observes or receives that change.

It turns the state into a passenger-safe message.

The message reaches screens, announcements, apps or staff.

Passengers change routes.

Their new movement changes station and network loads.

The information system must observe again.

railway state
→ public message
→ passenger decision
→ passenger flow
→ new railway state
→ revised message

MRT passenger information works when the right truth reaches the right passenger early enough to preserve a useful next move.

Key Equations

τ=treceived−tmeasured
Information age

estale≈(dx/dt)τ
Stale-state error scale

Tarrival=tnow+future dwell+run+control delay
Arrival prediction structure

eETA=Tactual−Tpredicted
Arrival prediction error

MAE=(1/n)Σ|eETA|
Mean absolute ETA error

P(A)=e^(−θCA)/[e^(−θCA)+e^(−θCB)]
Simple route-choice probability

Utility(message)=decision improvement−cognitive load−ambiguity−latency cost
Conceptual message utility

L=received-information journey cost−best-information journey cost
Decision loss

Reader-safety note: This article explains passenger-facing information and public communications. It does not reproduce restricted railway communications architecture, signalling protocols, cybersecurity controls, emergency command channels, internal operating codes or security-sensitive network details.

How MRT Communications and Passenger Information Work Using Mathematics: Why Correct Information Can Still Be Wrong If It Arrives Too Late