There is a particular kind of Singapore miracle that happens so routinely that most of us stop noticing it.
You walk onto an MRT platform. A display says the next train is arriving. A few moments later, headlights appear down the tunnel. The train slows, reaches the platform, lines itself up with the doors, stops, opens, exchanges hundreds of passengers, closes its doors and disappears towards the next station.
Then another train does it. And another.
Singapore’s rail network now spans more than 170 MRT stations and about 240 kilometres, carrying millions of journeys. The Land Transport Authority describes typical train frequencies as about two to three minutes during peak periods and about five to seven minutes off-peak. To the passenger, all of this feels ordinary. Mathematically, it is anything but ordinary.
An MRT railway is a gigantic moving timing problem. Every train has mass. Every train accelerates and brakes. Every train occupies physical space. Every passenger requires time to board and alight. Every station has a platform. Every junction has finite capacity. Every terminal must reverse trains. Every train ahead changes the safe operating envelope of the train behind it.
And all of these variables are moving at the same time.
This means the question, “How fast does an MRT train go?”, is actually one of the least interesting questions we can ask.
How does an MRT know how fast it should be moving right now so that the entire railway continues to work several stations — and several trains — later?
That question takes us into mathematics: arithmetic, algebra, geometry, kinematics, calculus, differential equations, statistics, probability, optimisation, queueing theory, graph theory, operations research and control theory.
The RFE — Reason for Existence
Why does all this mathematics exist?
Not to make one train move as fast as possible.
The railway exists to move people. So the real mathematical objective is closer to:
Move many people + move them safely + move them regularly + move them predictably + recover when something goes wrong
while using a finite number of trains, tracks, stations, electrical systems and seconds. That changes everything.
Prompt 1 — What Is an MRT Actually Timing?
Imagine looking at an MRT journey through the eyes of a passenger. We might describe it as:
Wait → Board → Travel → Stop → Alight
Simple.
Now look at the same journey from inside the railway. The system has to consider something more like:
Passenger demand ↓ Required service frequency ↓ Train headway ↓ Departure timetable ↓ Safe movement authority ↓ Permitted speed ↓ Acceleration / cruising / coasting / braking ↓ Station arrival ↓ Door opening ↓ Passenger exchange ↓ Door closure ↓ Next departure
And then the calculation starts again. Except that it is not really starting again. Every new calculation inherits the consequences of everything that happened before it.
A train that leaves Station A five seconds late may reach Station B three seconds late. Or seven seconds late. Or recover the entire delay. Or cause the train behind it to slow. Or encounter a larger crowd because passengers have continued accumulating on the platform.
Therefore MRT timing is not merely:
Time = Distance ÷ Speed
It is a dynamic system.
The many clocks inside one railway
We can think of an MRT as having several mathematical clocks running simultaneously.
- The physics clock: how long physically does the train require to travel the distance?
- The safety clock: how closely may the train approach another train?
- The platform clock: how long will passengers take to leave and enter?
- The timetable clock: where should the train be relative to its planned position?
- The network clock: how does its movement interact with other trains, junctions and terminals?
- The energy clock: can the same journey be achieved using less traction energy?
- The recovery clock: if something has gone wrong, how much flexibility exists to restore normal operation?
The passenger sees one arrival time. The railway has to reconcile all of these clocks.
MRT timing is not one timer. It is coordination among many timers.
Prompt 2 — How Fast Can the Train Actually Travel Between Two Stations?
Now we can introduce the mathematics most students will recognise.
Suppose two hypothetical MRT stations are 1.2 kilometres apart:
1.2 km = 1,200 m
Suppose, purely for our worked example, that the train’s maximum permitted speed is 80 km/h:
80 km/h = 22.22 m/s
The 80 km/h figure is a useful Singapore reference point. LTA has documented the return of maximum speeds of 80 km/h on the North-South and East-West Lines after earlier temporary speed restrictions connected with sleeper replacement works were removed.
But here is the mistake we must avoid. We cannot simply calculate:
T = 1,200 ÷ 22.22 ≈ 54 seconds
and conclude that the journey takes 54 seconds, because the train does not magically begin its journey already travelling at 80 km/h.
At the station, velocity is zero. The train must accelerate. Before the next station, its velocity must return to zero. So the journey consists of phases:
Accelerate → Cruise → Brake
A more energy-conscious trajectory may instead resemble:
Accelerate → Cruise → Coast → Brake
Acceleration
Suppose our illustrative train accelerates at 1.0 m/s². The familiar kinematics equation is:
v² = u² + 2as
With initial velocity u = 0:
sₐ = v² ÷ 2a sₐ = 22.22² ÷ 2 sₐ ≈ 247 m
The time required is:
tₐ = (v − u) ÷ a ≈ 22.2 s
Already, nearly 250 metres of our 1.2-kilometre journey have disappeared just getting up to speed.
Braking
Now suppose our simplified braking deceleration is 1.1 m/s². Stopping distance becomes:
sᵦ = v² ÷ 2b ≈ 224 m
and braking time is approximately:
tᵦ = v ÷ b ≈ 20.2 s
Acceleration and braking together use about:
247 + 224 = 471 m
That leaves:
1,200 − 471 = 729 m
for cruising. At 22.22 m/s, that takes approximately 32.8 seconds. Total simplified running time becomes:
22.2 + 32.8 + 20.2 ≈ 75.2 seconds
That is already very different from our naive 54-second answer.
But even this is still too simple
A real train trajectory does not necessarily look like a perfect textbook trapezium. We should think of velocity and position as functions:
v = v(t) x = x(t) v(t) = dx/dt a(t) = dv/dt
Now calculus enters the railway. Distance is the integral of velocity over time:
x(t) = ∫ v(t) dt
The area under a velocity-time graph is the distance travelled. The train’s journey might therefore involve strong acceleration, reduced acceleration, a permitted top speed, cruising, coasting, a lower speed region, braking, final approach and a precise stop.
The timetable therefore does not need only a destination. It needs a trajectory.
Jerk: why acceleration itself cannot change carelessly
Passenger comfort introduces another derivative. Acceleration is:
a = dv/dt
but the rate at which acceleration changes is called jerk:
j = da/dt
If acceleration or braking changes too abruptly, standing passengers feel the change. So a passenger railway cannot optimise only for minimum journey time. It also needs a reasonably smooth acceleration profile.
RFE: Why not simply accelerate harder? Because the passenger is not freight. The railway is not trying to win a 1.2-kilometre sprint. It is trying to produce a journey that can be repeated safely and comfortably thousands of times.
Prompt 3 — What Forces Decide the Speed?
The train is not merely a point moving along a line. It is a physical body.
A useful simplified force equation is:
mₑff (dv/dt) = Fₜ − Fᵦ − R(v) − mg sin θ
where effective mass, traction force, braking force, resistance and track gradient all contribute to what the train can do at a given moment.
Mass matters
A lightly loaded train does not have exactly the same mass as a heavily loaded peak-hour train. Newton gives us:
F = ma a = F/m
For a fixed force, increasing mass decreases acceleration. Real train control therefore works with the physical characteristics and state of the train rather than pretending that every journey occurs in an empty mathematical universe.
Gradient matters
If the train travels uphill, gravity contributes a component opposing motion:
Fᵍ = mg sin θ
If the train travels downhill, gravity assists motion. Therefore:
One kilometre uphill is not mathematically equivalent to one kilometre downhill.
Both contain exactly 1,000 metres. They do not necessarily require the same energy or control trajectory.
Curves and infrastructure matter
Track curvature, station approaches, junctions, crossovers and other infrastructure can impose local speed limits. So permitted speed is better thought of as a function of position:
vₘₐₓ = vₘₐₓ(x)
The train may be mechanically capable of one speed while the infrastructure permits another. The operating requirement is therefore:
v(x) ≤ vₘₐₓ(x)
Resistance matters
Railway resistance is often represented with relationships of the general form:
R(v) = A + Bv + Cv²
The exact constants depend on the train and conditions. The important lesson is that resistance is not constant, and aerodynamic effects grow increasingly important at higher velocities.
Maximum speed is not the same as useful speed
If the next station is close, accelerating all the way to maximum speed only to brake immediately may save surprisingly little time while consuming disproportionately more energy. A slightly different trajectory may arrive only a few seconds later while using less traction energy.
Journey time ↔ Energy ↔ Comfort ↔ Safety ↔ Reliability
This is why “make the MRT faster” is not a complete engineering instruction.
Prompt 4 — How Does the Train Know How Close It Can Get to the Train Ahead?
Imagine Train A travelling ahead and Train B following. Train B cannot simply obey the track speed limit. Even if the railway allows a high speed at its present location, Train B may have to travel more slowly because Train A occupies the railway ahead.
The governing question becomes: if something changes ahead, can Train B still stop safely?
A school-level approximation of stopping distance is:
d = v² ÷ 2b
or, rearranged:
v = √(2bd)
The real railway calculation is much more sophisticated because signalling must account for safety margins, braking characteristics, gradients, position uncertainty and equipment behaviour. But the intuition is powerful:
The less safe stopping distance remaining, the lower the allowable speed.
Fixed blocks and moving information
Traditional railway signalling commonly divides lines into fixed sections or blocks. If a train occupies a block, following trains are kept out of unsafe territory behind it.
Modern Communications-Based Train Control, or CBTC, provides much more continuous knowledge of train position and allows trains to operate closer together while maintaining their required safety envelope.
Singapore’s North-South and East-West Lines underwent a major resignalling programme from their earlier fixed-block system to CBTC. LTA described the newer signalling as enabling trains to run closer together, with design capability reducing intervals from about 120 seconds to as low as approximately 100 seconds.
This does not mean every NSEWL train departs exactly every 100 seconds. There is an important difference between technical capability and the actual operating timetable. LTA’s passenger information currently describes typical peak train frequencies as approximately two to three minutes.
Headway mathematics
The interval between successive trains is called headway. Let the departure time of train i be Dᵢ. Then:
Hᵢ = Dᵢ − Dᵢ₋₁
If trains depart every 120 seconds, the theoretical number passing one point per hour is:
q = 3,600 ÷ H q = 3,600 ÷ 120 = 30 trains/hour
At 100 seconds:
q = 3,600 ÷ 100 = 36 trains/hour
Reducing the interval from 120 to 100 seconds therefore changes theoretical throughput from 30 to 36 trains per hour — a 20 per cent increase.
Twenty seconds suddenly looks enormous.
But signalling is only one constraint
If signalling safely permits a 100-second interval, can the whole line indefinitely operate every 100 seconds? Not necessarily.
Hoperating ≥ max( Hsignal, Hstation, Hjunction, Hturnback, Hpower, Hfleet, Hrecovery )
The tightest constraint can govern the line. This is a central piece of infrastructure mathematics:
System capacity is often determined by the bottleneck.
Prompt 5 — The Station Problem: Why Doors Can Matter More Than Speed
Now we reach something beautiful. The hardest part of an MRT timing problem may not happen while the train is moving. It can happen when the train is standing still.
In November 2025, Singapore’s Ministry of Transport stated that MRT and LRT dwell times depend on operating conditions and passenger loading. Typical dwell times were reported at approximately 35–60 seconds at interchange stations and 30–50 seconds at non-interchange stations.
Think about what that means. Our simplified 1.2-kilometre journey took about 75.2 seconds of running time. Add a 40-second station dwell:
75.2 + 40 = 115.2 seconds
The train can spend a very substantial fraction of the station-to-station cycle not moving at all. This is why simply increasing maximum train speed may accomplish much less than expected.
What is dwell time?
Define Td as dwell time. A conceptual model might be:
Td = Tdoors + Talighting + Tboarding + Tclearance + Tmargin
A more detailed explanatory model might be:
Td = T₀ + αA + βB + γC
where A is the number of passengers alighting, B the number boarding, C a crowding/interference term, and the Greek letters represent parameters estimated from observations. This is not presented as LTA’s proprietary formula. It is a mathematical model showing how such a problem can be represented.
Why crowding becomes nonlinear
Ten passengers entering a relatively empty doorway may be straightforward. One hundred people trying to board while many others are simultaneously trying to alight is different. People interact, paths cross and doorways become contested space. Therefore dwell time need not increase in a simple straight line with passenger count.
This is where pedestrian-flow modelling, probability and queueing theory enter the railway.
Accessibility is part of the timing mathematics
Why not simply cut every station dwell by ten seconds? Because passengers are not identical particles. MOT has explicitly noted that dwell times must allow safe boarding and alighting, including for elderly commuters and people using wheelchairs or personal mobility aids.
The mathematically fastest possible passenger exchange is not automatically the socially correct one.
The railway exists for the population, not merely for the statistically quickest passenger. Accessibility therefore appears inside an engineering timing problem. A human value becomes a number of seconds.
Passenger arrivals create a queue
Suppose passengers arrive at a platform at an average rate λ people per second. With headway H, the approximate number accumulating between trains is:
N ≈ λH
If λ = 2 passengers per second and H = 180 seconds, then approximately 360 passengers arrive between trains. Reduce headway to 120 seconds and the corresponding number becomes about 240.
Frequency therefore affects more than waiting time. It changes platform crowding, boarding demand, train loading, dwell time and the subsequent timetable.
A simple stability condition
Suppose the arriving train has effective remaining boarding capacity K. For the platform queue not to grow indefinitely, a simplified condition is:
λH < K
If instead:
λH > K
then more passengers are arriving between trains than the next train can remove. A simplified queue evolution is:
Qₙ₊₁ = Qₙ + λH − K
If this remains positive, the queue grows. Queueing mathematics becomes a packed MRT platform.
Prompt 6 — How Does One Late Train Make Other Trains Late?
Imagine Train A reaches an interchange station. Its planned dwell is 40 seconds, but passenger loading means it actually remains for 55 seconds.
Delay introduced = 55 − 40 = 15 seconds
A passenger may think: fifteen seconds — so what?
But Train B is behind Train A. Train C is behind Train B. The railway behaves like a connected system.
Delay as an error signal
Let Ts be scheduled time and Ta actual time. Define timing error:
e = Ta − Ts
If e = 0, the train is precisely on schedule. If e = +15 s, it is 15 seconds late. The control problem is to move:
e(t) → 0
without compromising safety or destabilising the trains around it.
Recovery time
A robust timetable does not normally require trains to perform every run at their absolute physical minimum time. If the physical minimum interstation run were 70 seconds but the timetable allocated 75 seconds, there would be five seconds of running-time flexibility.
A slightly late train may be able to use part of that flexibility, within all permitted limits, to reduce its delay. This is recovery margin.
A railway operated permanently at every theoretical limit might be very efficient when nothing goes wrong and very fragile when anything does.
Efficiency ↔ Robustness
The dangerous feedback loop
Suppose Train A is delayed. Its gap from the train ahead becomes larger. More passengers accumulate before Train A arrives. More passengers can mean a longer dwell. Longer dwell can produce more delay.
Delay → larger passenger accumulation → longer dwell → more delay
Meanwhile Train B may be catching up behind. Because Train A collected many of the passengers, Train B may encounter fewer people and a shorter dwell. Train A gets later while Train B catches up.
This is a form of bunching. Mathematically, it is positive feedback.
A simple recurrence
Let the delay of a train after station k be δk. A conceptual recurrence is:
δₖ₊₁ = δₖ + ΔTdwell,k + ΔTrun,k − Rₖ
where excess dwell and excess running time add delay, while Rk represents recovered time. If recovery exceeds the new disturbance, delay shrinks. If not, delay grows.
Now multiply that equation across many trains and many stations. You have a dynamic railway.
Control theory appears
This is why modern metro operation is not simply a timetable printed in the morning and followed blindly until midnight. Passenger loads, running times and operating conditions vary. Research into high-density metro operations therefore treats dwell-time control, timetable behaviour and train speed profiles as coupled optimisation problems.
The hidden mathematical character of an MRT is that it is constantly trying to keep a moving system inside an acceptable operating envelope.
Prompt 7 — Why Doesn’t Every MRT Train Simply Drive as Fast as Possible?
Because time is not the only thing being optimised.
Suppose two driving strategies both reach the next station on time.
Strategy A
Accelerate aggressively, maintain a high speed and brake later.
Strategy B
Accelerate appropriately, reach a suitable speed, coast for part of the journey and brake smoothly.
If both arrive inside the required timetable, Strategy B may use less energy. Many possible speed trajectories can satisfy the same arrival time:
Tarrival = Trequired
Coasting
Coasting means the train continues moving while traction power is not being applied and the brakes are not being used. Momentum keeps it moving.
SMRT has publicly described this idea in its Green CBTC programme for the North-South and East-West Lines. Its Automatic Train Operation system manages acceleration, braking and coasting, with algorithms following optimised running curves. SMRT describes predetermined coasting locations as a way to reduce energy consumption with negligible effect on journey time.
Energy as an integral
Mechanical power can be written as:
P = Fv
Energy over time is:
E = ∫ P(t) dt = ∫ F(t)v(t) dt
Two speed profiles can cover the same distance in almost the same time while producing different energy consumption.
A simplified optimisation objective might be:
Minimise J J = w₁T + w₂E + w₃D + w₄C + w₅R
where the terms can represent journey time, energy consumption, delay, discomfort and robustness or risk, with weights reflecting operating priorities.
The optimisation is constrained by requirements such as:
v(x) ≤ vmax(x) aₘᵢₙ ≤ a(t) ≤ aₘₐₓ |j(t)| ≤ jₘₐₓ H(t) ≥ Hₘᵢₙ
plus safe stopping, station positioning, movement authority and timetable constraints.
Singapore provides a real example
SMRT reported in November 2024 that the first phase of its Green CBTC work achieved an additional 8 per cent traction-energy saving, corresponding at that time to more than 15 million kWh annually. The programme demonstrates that the goal is not simply “slow trains” or “fast trains”. It is the appropriate trajectory.
Regenerative braking adds another layer
Electric trains can recover useful electrical energy during braking rather than converting all motion into wasted heat. That creates another intriguing optimisation problem.
If one train is braking while another is accelerating, there may be opportunities — depending on the traction-power system and operating conditions — to make useful use of regenerated energy. Academic research on metro timetable and speed-profile optimisation explicitly studies the synchronisation of acceleration and braking to improve regenerative-energy utilisation.
Now timing has become electrical coordination. One second can matter twice: once for punctuality, and once for energy.
Prompt 8 — The Whole Railway: From One Train to a City
We can now assemble the pieces.
A train does not really have an independent timetable. It occupies a place in a line. The line occupies a place in a network. The network serves a city.
Fleet mathematics
Suppose a complete round trip, including terminal turnaround and operating allowances, takes 120 minutes. If trains are required every two minutes, an approximate fleet requirement in the active cycle is:
N ≈ Cycle time ÷ Headway N ≈ 120 ÷ 2 = 60 trains
Now suppose the target headway becomes 100 seconds:
100 s = 1.667 min N ≈ 120 ÷ 1.667 ≈ 72 trains
This is deliberately simplified and excludes spare trains, maintenance needs and many operating details. But it demonstrates an important fact: reducing headway may require substantially more rolling stock.
Therefore “run the trains more frequently” may imply more trains, more maintenance, more depot capacity, greater electrical demand, tighter terminal operation, more platform throughput and more resilience engineering.
Terminal mathematics
Eventually a train reaches the end of the line. It cannot continue forever. It must turn around.
If a terminal can process a reversing train only every HT seconds, then even a signalling system elsewhere capable of smaller headways cannot push more trains through that terminal indefinitely.
Hline ≥ HT when the terminal is the bottleneck
Singapore provides a concrete example. LTA completed the Pasir Ris rail turnback in April 2024. The new infrastructure allows trains to cross over more quickly and supports shorter East-West Line frequencies when needed.
Concrete tracks therefore alter an equation. Engineering changes Hturnback, and that changes possible line capacity.
Junction mathematics
Junctions create another scheduling problem. Imagine two streams of trains requiring the same piece of track. Train A wants it at time tA. Train B wants it at tB. They cannot occupy conflicting routes simultaneously.
|tA − tB| ≥ S
where S is the required separation. Across a large railway this resembles other operations-research problems: aircraft runway scheduling, manufacturing systems, computer processors, shipping terminals, telecommunications and logistics networks.
A railway is a physical computer made of trains.
Graph theory: the MRT as a network
Represent every station as a node V and every rail connection as an edge E. Then:
G = (V, E)
Each edge can carry values such as travel time, capacity, energy cost, reliability, speed limit and passenger volume. Transfer stations connect multiple paths.
Journey planning then becomes a path problem through a graph. The shortest-distance route need not be the shortest-time route. The shortest-time route need not minimise transfers. The route with the fewest transfers need not minimise expected delay.
Even the passenger’s trip is an optimisation problem.
Probability enters
Until now, many equations have looked deterministic. Real life is not.
Dwell time varies. Passenger demand varies. Equipment behaviour varies. Walking speed varies. Incidents occur. So variables such as Tdwell and Trun are often better understood statistically.
Instead of asking only for expected time E[T], a robust railway may care about:
P(T > Tlimit)
the probability that a journey or process exceeds an unacceptable threshold.
Two systems can have the same average journey time but radically different reliability. One may consistently take 29–31 minutes. Another may often take 25 minutes but sometimes take 45. The mean can be similar while passenger experience is not.
The timetable is a probability shield
A timetable contains allowances not because engineers are unable to calculate minimum journey time, but because reality varies. A strong timetable tries to absorb ordinary disturbances without allowing them to become system-wide failures.
Robust timetable = efficient plan + carefully positioned flexibility
Too much margin wastes capacity and time. Too little creates fragility. The useful answer lies between them.
A Second-by-Second Illustrative MRT Journey
Let us combine everything into one simplified example. This is not an actual Singapore MRT operating profile. It is an educational reconstruction showing the mathematics an MRT-like system must solve.
Suppose Station A and Station B are 1,200 metres apart. The train is scheduled to depart at 16:45:00. Assume:
vmax = 22.22 m/s a = 1.0 m/s² b = 1.1 m/s²
16:45:00 — Departure
Position x = 0. Velocity v = 0. Doors are confirmed closed, movement is permitted and acceleration begins.
16:45:10
Using v = at:
v = 10 m/s = 36 km/h
Distance from s = ½at² is:
s = 50 m
The train has travelled only 50 metres.
Around 16:45:22
Velocity reaches approximately 22.22 m/s, or 80 km/h, after about 247 metres of acceleration in our deliberately simplified model.
Cruise or coast phase
Approximately 729 metres remain between our simplified acceleration and braking regions. At the assumed maximum speed that takes about 32.8 seconds. An energy-efficient real trajectory might include coasting and therefore use a different velocity curve while still respecting the required arrival time.
Approximately 16:45:55 — Braking begins
Why brake here? Because stopping is a distance problem. At 22.22 m/s with our assumed deceleration, stopping distance is about 224 metres. If braking waits too long, the remaining distance becomes insufficient.
The station is affecting the train hundreds of metres before arrival. The future controls the present.
Approximately 16:46:15 — Arrival
The train reaches v = 0 at Station B. Simplified running time is about 75.2 seconds. If actual dwell is 40 seconds, the doors may be ready for departure at roughly 16:46:55.
One hypothetical station-to-station cycle has therefore occupied about 115 seconds from departure at one station until departure from the next.
Now imagine this process repeated across dozens of stations, by dozens of trains, in both directions, with passengers constantly entering the system.
That is MRT timing.
What Happens If We Save Five Seconds?
Suppose engineers save five seconds on one interstation run. If a train crosses 25 such sections:
25 × 5 = 125 seconds = 2 minutes 5 seconds
But suppose those five seconds require aggressive acceleration, higher energy use and the removal of useful recovery margin. The system may become less robust.
Conversely, suppose five seconds are saved through better signalling, a faster terminal crossover, improved passenger distribution, more consistent dwell or better trajectory optimisation. Then the gain may propagate through the timetable.
Not all seconds are equal.
What Happens If We Lose Five Seconds?
Five seconds lost once is insignificant. Five seconds lost repeatedly is not.
If a delay increment d occurred across n stations in a purely additive model:
D = nd
For d = 5 seconds and n = 20:
D = 100 seconds
But railway delays are not simply additive. Delay changes passenger loads and following headways, so a more realistic conceptual relationship is:
Dₖ₊₁ = f(Dₖ, Pₖ, Hₖ, Rₖ)
Now we are looking at a nonlinear dynamical system.
Why MRT Mathematics Is a Beautiful Education Problem
A student learning mathematics often meets each topic separately.
- Algebra: solve for x.
- Kinematics: calculate acceleration.
- Calculus: differentiate x(t).
- Statistics: calculate a mean.
- Probability: find P(X > x).
- Graph theory: find a route through a network.
- Optimisation: find the minimum or maximum under constraints.
But an MRT does not care which school chapter these equations came from. Reality mixes them together.
Kinematics ↓ Calculus ↓ Control ↓ Probability ↓ Queueing ↓ Optimisation ↓ Human behaviour ↓ Time
That is what mathematics looks like after it leaves the examination paper.
RFE — What Is All This Mathematics Really For?
We can now return to the most important question: what is the Reason for Existence of MRT timing?
Imagine optimising only one variable:
Minimise journey time
Trains could accelerate harder. Margins could shrink. Dwell could be cut. Recovery time could disappear.
But passengers might experience uncomfortable acceleration. People requiring more boarding time could be disadvantaged. Energy demand could rise. Small disturbances could produce larger disruptions.
The result could be a mathematically fast railway and a practically poor railway.
So the real objective is multidimensional:
Safety + Capacity + Reliability + Punctuality + Accessibility + Energy efficiency
subject to limited track, trains, power, space, money, maintenance windows and time.
There is no single magic number. There is a balance.
The Faster Train Paradox
A railway can sometimes improve people’s journeys without making trains travel faster.
- Reduce irregular headways: passenger waiting becomes more predictable.
- Make dwell times more consistent: delay propagation decreases.
- Improve terminal turnaround: more train paths become usable.
- Improve signalling: trains can safely operate closer together.
- Improve passenger distribution along platforms: door congestion may decrease.
- Optimise acceleration and coasting: energy use can fall without materially changing travel time.
- Improve recovery: disturbances can disappear more quickly.
The passenger experiences a better railway while maximum train speed may remain exactly the same.
A faster railway is not necessarily a railway with faster trains. Sometimes it is a railway with better mathematics.
Why 100 Seconds Can Be More Interesting Than 80 km/h
The public tends to notice maximum speed, but for a dense urban railway, headway can matter even more.
Increasing top speed from 80 to 85 km/h might save surprisingly little on closely spaced stations because trains spend substantial time accelerating, braking and dwelling. But reducing a theoretical headway from 120 to 100 seconds changes flow from 30 to 36 trains per hour.
If each train carries a large number of people, those additional train paths represent substantial passenger-carrying capacity.
Urban rail therefore values frequency and regularity at least as much as raw top speed. This is one reason a metro is fundamentally different from a high-speed railway.
The MRT as a Moving Equation
We can now write a more complete conceptual model. For train i:
xᵢ(t) = position vᵢ(t) = dxᵢ/dt aᵢ(t) = dvᵢ/dt jᵢ(t) = daᵢ/dt
Movement must satisfy the local speed envelope:
vᵢ(t) ≤ vmax(xᵢ)
and safe separation:
xᵢ₋₁(t) − xᵢ(t) ≥ dsafe,i(t)
while trying to reach station k near its required time:
Tactual,i,k ≈ Tschedule,i,k
and maintain an appropriate headway:
Hᵢ = Tᵢ − Tᵢ₋₁
while dwell depends on passenger and operating conditions, passenger accumulation depends partly on headway, and energy depends on the entire speed and force history:
N ≈ λH E = ∫ F(t)v(t) dt
And the whole system is trying to reduce delay, energy use, headway variation, discomfort and risk while obeying physical and operating constraints.
That is an MRT: a moving collection of equations.
And Then the World Returns
There is one final step that textbooks often forget.
The railway makes a plan. Then reality answers.
The timetable expects 40 seconds of dwell. Reality gives 52. The model expects one passenger load. Reality gives another. Predicted running time is one value. Actual operating conditions produce another.
Predict → Act → Observe → Compare → Correct → Act again
Mathematically, the difference can be represented as an error:
e(t) = yobserved(t) − ydesired(t)
The system responds. Then it observes again. This is control.
The timetable is only a prediction of the future. The railway has to survive the actual future.
Conclusion — The Seconds Beneath Singapore
The next time you stand on an MRT platform, notice the seconds.
Watch a train enter the station. It is already braking because of mathematics performed before it reaches you.
Watch the doors open. Passenger movement is now part of the timetable.
Watch the doors close. A timer has become a capacity constraint.
Feel the train accelerate. Its speed trajectory is balancing time, traction, comfort, energy and the position of the train ahead.
Feel it coast. Doing apparently “nothing” may be the mathematically efficient decision.
Watch another train arrive a few minutes later. The interval between the two is headway. Headway determines how many trains can pass. That determines passenger capacity. Passenger capacity changes crowding. Crowding changes dwell. Dwell changes headway. Headway changes the next crowd.
Everything returns.
So the deepest equation in an MRT is not:
Speed = Distance ÷ Time
It is closer to:
People + Physics + Space + Time + Uncertainty + Control = A Working Railway
Singapore’s MRT does not work because every train is moving as fast as it can. It works when every train is moving as fast as it should, stopping for as long as it should, staying as far from the next train as it should, and returning towards the intended rhythm whenever reality pushes it away.
The passenger sees a train. The mathematician sees velocity. The engineer sees constraints. The controller sees error. The planner sees capacity. The city sees connectivity.
But underneath all of them is the same scarce resource:
Time.
And much of the art of running an MRT is deciding, second by second, where that time should go.
Key Mathematical Ideas Used in MRT Timing
v = dx/dt Velocity a = dv/dt Acceleration j = da/dt Jerk v² = u² + 2as Kinematics dstop ≈ v²/(2b) Simplified braking distance Hᵢ = Dᵢ − Dᵢ₋₁ Train headway q = 3,600/H Theoretical trains per hour N ≈ λH Passenger arrivals between trains Qₙ₊₁ = Qₙ + λH − K Simplified platform queue Nfleet ≈ Tcycle/H Simplified fleet requirement P = Fv Mechanical power E = ∫P(t)dt Energy e = Tactual − Tscheduled Timetable error e(t) → 0 A control objective
Good MRT timing is not minimum travel time. Good MRT timing is safe, regular, recoverable, accessible, energy-aware and high-capacity movement.
Sources and Further Reading
- Land Transport Authority — Rail Network and Train Frequency
- Ministry of Transport — Dwell Times for Trains at Interchange and Non-Interchange Stations
- Land Transport Master Plan 2013 — NSEWL Signalling Capacity Context
- LTA Annual Report 2023/24 — Pasir Ris Rail Turnback
- SMRT — Green CBTC Energy Savings on NSEWL
- SMRT — Automatic Train Operation, Coasting and Green CBTC
- Transportation Research Part B — Timetable and Speed-Profile Optimisation for Regenerative Energy
- Transportation Research Part B — Metro Dwell, Timetable and Speed-Profile Optimisation
Return to the complete MRT Mathematics hub: How MRT Works | It’s Mathematics.