The Nagel-Schreckenberg Traffic Model, Explained

After reading this you will understand why traffic jams form on an open road with no crash and no bottleneck, how four simple rules produce that behaviour, and how to read the flow curve to find the density that moves the most cars.

What the model is and why jams appear from nothing

The Nagel-Schreckenberg model is a cellular automaton for single-lane traffic. Introduced by Kai Nagel and Michael Schreckenberg in 1992, it divides a road into equal cells. Each cell is either empty or holds one car, and each car carries an integer speed from 0 to a maximum v_{max}. Time advances in discrete ticks. On every tick, every car applies the same four rules at once.

The striking result is that jams appear without any external cause. Put enough cars on a ring road, let one driver tap the brakes once, and that single slowdown can grow into a dense band of stopped cars that drifts backward against the flow of traffic. Nobody crashed. There is no on-ramp, no red light, no lane closure. The jam is a property of the collective, not of any one driver. On a real highway you meet this as the phantom jam: you crawl for two minutes, then the road clears and you never see a reason.

The simulator puts cars on a closed loop, so the number of cars is fixed. That lets you study one density at a time and watch the same population circulate.

When to use it, and when not to

Use this model to build intuition about three things: that congestion is an emergent phase, that flow does not rise forever with density, and that small noise can trigger large structure. It is the cleanest available demonstration that a traffic jam is a wave, not a place.

Do not use it to predict tomorrow's commute. The model has one lane, identical drivers, no junctions, and a single dawdle probability. Real traffic has overtaking, trucks, merges, and drivers who react differently. The model reproduces the shape of the fundamental diagram, not the exact numbers for a given road.

The four rules and the intuition behind each

Let car i have speed v_i and let d_i be the number of empty cells between it and the car ahead. One tick updates every car in this order, using the same old positions for all gap measurements.

1. Accelerate. If v_i \lt v_{max}, increase v_i by 1. Drivers want to go faster.
2. Brake. If v_i \gt d_i, set v_i = d_i. You cannot drive into the gap you do not have.
3. Dawdle. With probability p, if v_i \gt 0, decrease v_i by 1. This is the only randomness: a distraction, an overreaction, a slight tap of the brake.
4. Move. Advance the car v_i cells.

The braking rule can be written compactly. After acceleration the new speed is capped by the gap:

Here v_i + 1 is the wish to speed up, d_i is the safety limit set by the car ahead, and v_{max} is the hard cap. Then with probability p the model subtracts one more, giving the final speed \max(v_i' - 1,\; 0) on dawdle ticks. Without the dawdle (p = 0) the model is deterministic and, above a certain density, settles into smooth patterns. The noise is what makes jams appear and disperse the way real ones do.

Density, flow and the fundamental diagram

Two summary numbers describe the whole road. Density \rho is the fraction of cells that hold a car: with N cars on L cells, \rho = N/L. Flow q (also called throughput) is the average number of cars that cross any fixed point per tick. It equals density times mean speed:

In this equation \bar{v} is the average speed of all cars measured in cells per tick. The plot of q against \rho is the fundamental diagram, and its shape is the whole story. At low density cars rarely meet, so \bar{v} \approx v_{max} and flow climbs almost linearly. Past a critical density \rho_c the road is crowded enough that jams persist, mean speed falls faster than density rises, and flow drops. Add more cars beyond the peak and you move fewer of them past any point.

A worked example with the demo settings

The demo button loads the field defaults. Work through one tick by hand to see the rules bite, then reason about the whole road.

Now scale up. Set the density to the default and watch the whole ring. Below the critical density most cars sit at speed 5 and the space-time diagram shows nearly straight bright lines. Push density up and dark backward-slanting stripes appear. Each stripe is a jam. Its slope is the jam's velocity: in this model a well-developed jam drifts backward at close to one cell per tick, roughly the reciprocal of the dawdle interplay, regardless of how fast the free-flowing cars move.

Reading and interpreting the results

Three things on the screen carry the meaning. First, car colour by speed: a road of one bright colour is free flow, patches of dark are jams. Second, the space-time diagram, where the horizontal axis is position and each new row is a later tick. A jam shows as a dark band tilting up and to the left, because the stopped region sits at earlier positions as time passes. That leftward tilt is the backward wave. Third, the live flow number. Watch it settle after the first hundred or so ticks; early values reflect the starting layout, not the steady state.

To find the peak of the fundamental diagram, hold v_{max} and p fixed and raise density in small steps, recording the settled flow at each. With v_{max} = 5 and p = 0.3 you should find flow peaking near \rho = 0.16 at about 0.65 cars per tick, then declining. That peak density is the most efficient loading of the road.

Common mistakes when running it

A second mistake is expecting a sharp critical density. The transition from free flow to jams is gradual and noisy, not a clean edge. Near \rho_c you will see jams flicker in and out; that is real behaviour, not a bug.

A third is confusing high density with high flow. It feels like packing more cars should move more traffic. The diagram says otherwise: past the peak, every extra car lowers throughput. On real roads this is why ramp metering, which limits how fast cars join, can raise the total number that get through.

Finally, remember the model is one lane on a loop. There is no overtaking. A single slow car sets the pace for everyone behind it until a gap opens, which exaggerates how fast jams form compared with a multi-lane road.

Related simulations on this site

If emergent collective behaviour from simple local rules interests you, two other simulators here explore the same idea in different settings. The SIR Epidemic Simulator shows how an outbreak grows and burns out from three population rates, with its own critical threshold for whether a disease spreads at all. The Predator-Prey Simulator traces boom-and-bust cycles from coupled rate equations. All three share a lesson: rich, wave-like or cyclic patterns come from a handful of rules, not from any single driver, predator or patient acting alone.

Related tools

Gerrymandering Sandbox

Draw the districts yourself and watch the same voters produce opposite election results.

Pareto Frontier Evolution

Watch a genetic algorithm breed bridge designs and trace out the weight-vs-stiffness Pareto frontier, generation by generation.

SIR Epidemic Simulator

Watch an epidemic spread and burn out with the classic susceptible–infected–recovered model.

Sailing & Tacking Simulator

Sail a boat against the wind — trim to the polar diagram, ride the shifts and race the optimal route.

Schelling’s Segregation Model

Give agents a mild preference for similar neighbours and watch a mixed city sort itself into stark blocks.

Tsunami Simulator

Set off an undersea earthquake anywhere and watch the tsunami cross real ocean floors — with arrival times at tide gauges around the world.