Delay at a junction is waiting caused by conflicts: at a roundabout and a priority junction a driver must find a gap in the stream it has to give way to; at a signalised junction the signal removes the conflict but makes everyone wait through the red. All three are compared at the same demand with the formulas of the Highway Capacity Manual family and then checked against an explicit vehicle-by-vehicle simulation.
gap acceptance (Tanner): c = v_c e^(-v_c t_c) / (1 - e^(-v_c t_f)) v_c = conflicting flow
roundabout entry (Siegloch, HCM 6 single lane): c = 1380 exp(-1.02e-3 v_c) t_c = 4.98 s, t_f = 2.61 s
circulating flow past entry j: v_c,j = sum of q_i f_k over the movements that pass entry j (exits at j leave first)
priority minor road: t_c = 6.2 / 6.5 / 7.1 s (right / through / left), t_f = 3.3 / 4.0 / 3.5 s, lane shared harmonically
signals: Webster C0 = (1.5 L + 5)/(1 - Y), g_i = (C - L) y_i / Y, y = q/s, L = 8 s lost per cycle
uniform delay d1 = C (1 - g/C)^2 / (2 (1 - x g/C)), x = q C/(s g); d2 = 900 T [(x-1) + sqrt((x-1)^2 + 8 k x/(c T))]
control delay d = 3600/c + 900 T[(x-1) + sqrt((x-1)^2 + (3600/c) x/(450 T))] + 5 (gap acceptance, T = 1 h)
queue (Little) N = q d / 3600
- Where each type wins - the roundabout has almost no delay at low flow (no red to wait through) but its capacity falls exponentially with circulating flow; signals cost a few seconds even in light traffic but keep their capacity; a priority junction is cheapest and fine while the main road is quiet, and collapses for the side road long before the main road is busy.
- Webster cycle - a short cycle wastes time in lost start-up and clearance, a long one makes people wait through a long red; the optimum balances the two, and the check scans the cycle by brute force to confirm it.
- Microsimulation - cars with a Gipps-type safe-speed car-following rule, a 1 s start-up reaction, Poisson arrivals, gap acceptance at yield lines and a fixed-time two-phase signal. Delay is measured against a solo run of the same vehicle through the same geometry, so only waiting caused by other traffic or by the signal counts. The model critical gap is the HCM value less 3 s because the 1 s start-up reaction, the acceleration from the line and the clearance of the conflict area are simulated explicitly.
- Same model in both pages - one shared physics module, only the drawing differs; the checks run three independent 1 h simulations with different random seeds.
- Honest simplifications - one lane per approach and per ring, 20 % right turns, no pedestrians, buses, bicycles or heavy vehicles (cars and vans only, pce 1), identical drivers, symmetric geometry, no signal actuation, left turns at the signals are permissive (they yield to oncoming traffic and may wait in the junction), side-road streams of a priority junction do not conflict with each other, the formulas ignore platooning, and results near or above capacity depend strongly on the hour of observation. Not a design tool.