Traffic Intersection Simulator

What you just saw

Vehicles arrive at each approach as a Poisson process, queue at the stop bar, and clear when the light turns green. When arrival rate exceeds the capacity of a phase, queues grow without bound — the classic failure mode of under-timed signals. The average wait time is sensitive to both arrival rate and green-phase duration, and the optimum depends on both approaches together.

The science behind it

Microscopic traffic simulation dates back to the 1950s. The per-lane Poisson arrival + fixed-time signal model here is the simplest useful version, underlying the Webster (1958) delay formula still taught in transportation engineering. Modern operational tools (VISSIM, AIMSUN, SUMO) add car-following dynamics, lane-changing, and turning movements, but the queue-growth intuition from this model carries over.

Try these experiments

1. 1. Balanced demand

   Settings: arrival_rate_ns=0.2 arrival_rate_ew=0.2

   What to look for: Both queues clear each cycle. Average wait sits near the half-cycle time.

2. 2. Mismatched signal

   Settings: arrival_rate_ns=0.4 arrival_rate_ew=0.1 green_duration_ns=20

   What to look for: NS queue grows without bound. Classic signal mis-timing failure mode.

3. 3. Rush hour

   Settings: arrival_rate_ns=0.6 arrival_rate_ew=0.5

   What to look for: Both approaches saturate. No signal timing makes the intersection work — you need more capacity.

from scirouter import SciRouter
client = SciRouter(api_key="sk-sci-...")
result = client.simulation.run(
    model="intersection",
    params={"arrival_rate_ns": 0.3, "arrival_rate_ew": 0.2,
            "green_duration_ns": 30, "green_duration_ew": 25},
    steps=3600, seed=42,
)
print(result.metrics["average_wait"])