Drishti

Urban Traffic Networks

Asphalt graphs where cars are packets you can't drop

A grid where cars are packets you can't drop when the queue backs up

At a glance

Flows
Vehicles, pedestrians, transit, emissions
Optimizes
Commute time, throughput, safety
Persists
Lane capacity limits and century-old street geometry
Likely future
Dynamic tolling and coordinated autonomous platoons

Seven lenses on Urban Traffic Networks

Same order every time. Expand the lens you need; first is open by default.

◎ What Exists — What is this, really?
What Exists

A jam is a density wave, not an object you can tow away.

Roads, lights, and intersections are the graph. Cars are packets that occupy real space — no buffer, no drop.

Zoning that separates sleep from work forces daily migrations that stress the grid. You cannot point to a jam; only to cars inside it.

↻ What Changes — What is stationary vs non-stationary?
What Changes

Rush hour is a clock-driven spike; rain and games are surprise injections.

Nine-to-five arrivals turn a free highway into stop-and-go. Rain widens following distance and cuts lane capacity instantly.

A stadium emptying is a localized packet flood. Rush hour slowly widens from one hour to three as drivers shift earlier or later.

→ What Flows — What moves? Where are bottlenecks?
What Flows

Flow dies at the slowest intersection, not the fastest lane.

flowchart LR
  Hwy[Highway lanes] --> Merge[Merge point]
  Merge --> Light[Signal intersection]
  Light --> City[City grid]

On-ramps exceed merge rate and queues spill backward. Gridlock: cars block intersections because exits are full — flow hits zero on intact asphalt. Queueing theory: speed limits barely matter at peak; clearing rate at bottlenecks does.

◈ What Learns — What updates, remembers, optimizes?
What Learns

Drivers learn shortcuts; signals learn timing — often fighting each other.

Navigation apps rebalance traffic onto quiet streets that immediately clog. Loop sensors let signals extend green for the longest queue — adaptive load balancing for steel. Waze may dump a thousand cars on a residential edge the city never planned for.

▣ What Persists — What survives change?
What Persists

One lane clears about 2,000 cars per hour — physics, not policy.

That cap persists regardless of speed limit or driver skill — reaction time and braking distance set it. Push following distance too low and free flow collapses into stop-and-go. Nineteenth-century street geometry still routes twenty-first-century traffic.

✦ What Emerges — How do simple rules become complexity?
What Emerges

Personal routing apps create commons tragedies and phantom jams.

Everyone optimizes alone; residential streets absorb highway overflow. Braess’s Paradox: a new shortcut can raise everyone’s commute by shifting equilibrium. One brake tap amplifies backward through dense traffic — a standing wave with no wreck at the front.

? What Will Happen — Which futures are becoming likely?
What Will Happen

Branches: platooning, priced roads, or worse gridlock.

BranchLeading indicator
AV platooning gains densitypermitted following distance in pilots
Congestion pricing worksvehicle-miles traveled in tolled zones
Sprawl wins politicallylane-miles added without VMT drop

Watch peak intersection clearing rate and toll-zone legislation — not ribbon-cutting on new highways.

Learn the mechanisms