🚦 Traffic: The Science of Congestion and Flow

What you’ll learn

1. The Jam With Nothing In ItRecognise that the everyday 'queue behind an obstacle' model of congestion makes three predictions that are all false, and accept the course's thesis in advance: congestion is a phase transition, not a queue.Almost everyone has crawled through a long jam and emerged to find nothing at its head — an experience the queue model cannot explain, because it predicts every jam has a cause, sits still, and can be cured with capacity. All three predictions fail: jams form with no obstacle at all, they travel upstream at a measurable speed, and added capacity fills. The alternative is that free-flowing and jammed traffic are two distinct phases of the same system, and above a critical density free flow is unstable in the technical sense — any tiny disturbance grows rather than fades, and somebody always taps a brake.
2. Three Numbers Describe All TrafficLearn the three measurable quantities of traffic — flow, density and speed — and understand why the identity q = k·v makes maximising throughput a genuine conflict rather than an addition problem.Flow (q) is vehicles past a point per hour, density (k) is vehicles per unit length of road, and speed (v) is how fast they're going; all three are directly measurable with a stopwatch and a bridge. They are locked by the identity q = k·v, which is arithmetic rather than a model. What makes it powerful is what it forbids: flow is what a road is for and it needs both terms high, but density and speed are not independent — packing cars in tighter forces drivers to slow, because the gap ahead is their safety margin. The whole subject is the question of what the product does when the two terms fight.
3. The Fundamental DiagramRead the fundamental diagram of traffic flow and take from it the three results that make traffic counterintuitive: capacity is a summit rather than a ceiling, flow collapses past critical density, and one flow value describes two entirely different phases.Plotting flow against density gives a hill: flow climbs as cars are added, peaks at the critical density, and then falls, so adding a car to a road past its critical density makes the road deliver fewer vehicles per hour than before — unlike a shop queue, where a new customer never slows the till. The Highway Capacity Manual has used a figure of about 2,000 passenger cars per hour per lane under ideal conditions since 1950, corresponding to an average speed of about 30 mph at about 67 pcpmpl, meaning a lane is at its most productive when it feels mediocre; modern estimates revise the number upward toward 2,000–2,200 but leave the shape untouched. Because the curve is a hill, almost every flow occurs twice — a sensor reading 1,500 veh/h cannot distinguish spread-out cars at 110 km/h from packed cars crawling — and the capacity drop means a road that has broken down recovers at a flow below its pre-breakdown capacity.
4. Sugiyama's CircleUnderstand Sugiyama's 2008 circular-track experiment as a decisive, filmed demonstration that jams form spontaneously above a critical density with no bottleneck at all — and that the 22-car threshold makes this a genuine phase transition.A team led by Yuki Sugiyama put 22 vehicles equidistantly on a uniform 230-metre circular track and asked drivers to hold a fixed speed and a fixed gap — a setup with no bottleneck, no junction, no obstacle, and no reason for anyone ever to brake. Within minutes a jam formed anyway: a cluster where cars came to a near-stop while cars elsewhere on the same ring drove freely, because no human holds a gap exactly, and each driver's slight over-correction amplifies the disturbance backwards down the line. The threshold was 22 cars — with 22 or more, bunches grew into a jam; with fewer, they resolved themselves — a measured critical point on a real road, supporting the authors' conclusion that a bottleneck is only a trigger and not the essential origin of a jam.
5. The Wave Goes BackwardsUnderstand a traffic jam as a wave with a measurable, near-constant upstream velocity independent of the cars composing it — and draw the practical consequences for lane-changing and for the jam you found nothing at the head of.A jam's population is continuously replaced — cars leave at the front and arrive at the rear — while the structure persists, has a length, and has a velocity that is not the velocity of any car in it, exactly as a sound wave crosses a room while the air molecules stay put. Because every car joins at the back, that structure travels upstream at a strikingly constant 15–20 km/h (c = −15 ± 5 km/h, established by Kerner and Rehborn in 1996) and, remarkably, without spreading, which is why researchers call it a traffic constant set essentially by driver reaction time and vehicle length. It follows that a jam formed at 8 a.m. is roughly 15 km further back down the road by 9 a.m. — it has outlived its own cause and left the scene — and that lane-changing cannot escape a wide wave while it does add a perturbation to a system defined by its amplification of perturbations.
6. Braess's ParadoxUnderstand Braess's paradox — that adding a road can make every driver slower — as a stable Nash equilibrium rather than a mistake, and see why real cities have found that removing roads sometimes speeds traffic up.Dietrich Braess showed in 1968 that 'adding one or more roads to a road network can slow down overall traffic flow through it': in the classic example, 4,000 drivers split evenly across two mirror-image routes each take 65 minutes, and adding a free zero-cost connector moves the equilibrium to 80 minutes for everybody. Nobody is irrational — the shortcut genuinely is faster given what everyone else is doing, but everyone reasons identically and switches onto the congestible legs, and the resulting bad state is stable because any driver returning to their old route is worse off still. The Nash equilibrium is not the system optimum, and cities have run the reverse experiment: Seoul's demolition of the Cheonggye Expressway in 2002, New York's Earth Day closure of 42nd Street in 1990, and Stuttgart in 1969 all improved conditions — though real cities are not controlled experiments and each closure came with other changes, so the pattern rhymes rather than proves.
7. Induced DemandUnderstand induced demand and the fundamental law of road congestion: that vehicle travel rises roughly proportionately with lane-kilometres, why the extra traffic appears, and why the finding extends to public transit.Duranton and Turner's 'The Fundamental Law of Road Congestion: Evidence from US Cities' (American Economic Review, October 2011), using US metropolitan data spanning roughly 1983–2003, found that vehicle-kilometres travelled increases proportionately to roadway lane kilometres for interstate highways and probably slightly less rapidly for other roads — build 10% more road, get about 10% more driving. The extra traffic comes from existing residents driving more, increased commercial traffic, and migration, none of which is a failure: congestion is a price paid in time, and cutting that price means people buy more travel. Their conclusion that increased provision of roads or public transit is unlikely to relieve congestion follows from congestion being an equilibrium set by drivers' tolerance rather than by capacity — which makes new capacity a genuine throughput gain, just not the congestion gain it is usually sold as.
8. What Actually WorksReframe congestion as a density-and-stability problem rather than a space problem, understand why ramp metering, congestion pricing and variable speed limits work while capacity does not, and generalise the failure of the container model.The course's seven prior results converge: throughput collapses past critical density, jams form with no obstacle, waves outlive their causes, added roads can slow everyone, and added capacity fills — so congestion is a density-and-stability problem and the levers that work manage those variables. Ramp metering rations entry to hold the mainline below critical density and avoid the capacity drop; congestion pricing changes the equilibrium by adjusting demand, at the cost of converting a burden paid in time into one paid in money, which is a real argument about who a city is for rather than a misunderstanding; variable speed limits damp disturbances upstream before they grow. Every counterintuitive result in the course comes from one mistake in different costumes — modelling responding agents as a passive container being filled — and that geometry of critical density, unstable phases, waves and self-filling capacity recurs anywhere individually sensible choices aggregate into a collectively bad outcome.

Questions this course answers

Grounded in trusted sources

• Sugiyama, Y., Fukui, M., Kikuchi, M., Hasebe, K., Nakayama, A., Nishinari, K., Tadaki, S. & Yukawa, S. (2008). 'Traffic jams without bottlenecks — experimental evidence for the physical mechanism of the formation of a jam.' New Journal of Physics 10(3), 033001. 230 m circular road, 22 vehicles equidistant, drivers told to hold fixed distance and velocity; critical threshold 22 vehicles; free flow destabilised by enhancement of fluctuations; 'a bottleneck is only a trigger and not the essential origin of a traffic jam'. https://iopscience.iop.org/article/10.1088/1367-2630/10/3/033001
• Duranton, G. & Turner, M. A. (2011). 'The Fundamental Law of Road Congestion: Evidence from US Cities.' American Economic Review 101(6), 2616–52. VKT rises proportionately with lane-kilometres for interstates; extra VKT comes from current residents, commercial traffic and migration; 'increased provision of roads or public transit is unlikely to relieve congestion'. https://www.aeaweb.org/articles?id=10.1257/aer.101.6.2616
• Wikipedia — "Braess's paradox": 'Adding one or more roads to a road network can slow down overall traffic flow through it'; 4,000 drivers, 65 minutes → 80 minutes after a zero-cost connector; Braess 1968, Pigou 1920; Seoul 2002, New York 1990, Stuttgart 1969. https://en.wikipedia.org/wiki/Braess%27s_paradox
• Wikipedia — 'Traffic flow': q = k·v; flow, density and speed definitions; the fundamental diagram; critical and jam density (185–250 veh/mile/lane, ~5× critical); regimes free-flowing <12, stable 12–30, unstable >30, breakdown >67 veh/mile/lane. https://en.wikipedia.org/wiki/Traffic_flow
• Kerner, B. S. & Rehborn, H. (1996), as summarised in Treiber, Kesting & Helbing, 'Empirical Features of Congested Traffic States and Their Implications for Traffic Modeling': the downstream front of a wide moving jam propagates upstream at a nearly constant 15–20 km/h (c = −15 ± 5 km/h) without spreading; typical national 'traffic constant' values 15 ± 5 km/h. https://arxiv.org/abs/cond-mat/0408138
• 'Freeway Capacity in Texas' (Summary Report 1196-2S, Texas Transportation Institute): the HCM has used ~2,000 passenger cars per hour per lane since 1950; under ideal conditions this corresponds to an average travel speed of 30 mph at a density of 67 pcpmpl; field flows routinely exceed the value. https://library.ctr.utexas.edu/hostedpdfs/tti/1196-2s.pdf
• MIT Mathematics — 'Phantom Traffic Jams and Traveling Jamitons': jamitons travel backwards on the road and slower than the individual vehicles; MIT simulations of the 230 m ring matched Sugiyama's experimental result. https://math.mit.edu/traffic/

Every Wunder lesson is built from real, reputable sources — never invented.

Related Science courses