Night-time signal timing is one of the most under-resourced areas of intersection design in Australia. Most signalised intersections run two or three timing plans: peak, inter-peak, and sometimes a weekend plan. After midnight, the controller often defaults to the inter-peak plan unchanged, or drops to a fixed minimum-cycle plan that was configured during commissioning and hasn't been reviewed since. That approach works until it doesn't. Low-volume night conditions are genuinely different from daytime inter-peak, and treating them identically creates measurable problems for freight, emergency services, and the pedestrians still using the network at 2 am.
What makes night-time traffic conditions distinct
Volume is the obvious difference. But volume alone doesn't explain why night-time plans need separate engineering. The composition of traffic changes significantly after midnight. Freight vehicles make up a larger share of movements on arterial roads during off-peak hours, and they have longer start-up lost times and wider turning templates than passenger cars. A cycle length optimised for cars under moderate volume can create excessive delay for heavy vehicles even when the intersection is lightly loaded overall.
Pedestrian demand doesn't disappear at night. It concentrates. Entertainment precincts, hospitals, late-shift workers, and food and drink venues generate pedestrian crossings in short, clustered bursts between 11 pm and 3 am. A controller running low minimum green times to keep vehicle throughput high will clip pedestrian crossing time precisely when the pedestrians using those crossings have reduced reaction speeds or reduced familiarity with the intersection. Pedestrian signal timing in high-density areas already demands careful calibration during the day; at night, the risk profile shifts further toward safety rather than throughput.
Emergency vehicle response times are also more sensitive to signal timing at night. With less traffic naturally clearing intersections, an ambulance travelling a 5 km corridor can accumulate substantial delay at red phases that would be cleared by vehicle actuation during the day. Signal plans that extend maximum green times to reduce cycle frequency create longer wait gaps for cross-traffic, which is fine when cross-traffic volumes justify the cycle. At low volumes, those same long cycles become obstacles.
Common problems with default off-peak plans
The most frequent issue is cycle length. Controllers left on inter-peak plans at night often run 90-second or 120-second cycles at intersections where traffic volumes would justify 60 seconds or less. A driver approaching a red signal at 1 am on an otherwise empty arterial, waiting 90 seconds for a green that serves two vehicles on the cross-street, isn't just inconvenient. It increases the likelihood of red-light running. Studies on isolated night-time intersections consistently show that long fixed cycles with low cross-street demand are a contributing factor in red-signal violations.
Pedestrian timing is the second problem. Minimum pedestrian green times are often set to 7 or 10 seconds in inter-peak plans, calibrated for a mix of able-bodied adults during daytime. Night-time users include people leaving licensed venues, older pedestrians, and hospital visitors, all of whom may need longer crossing times. The relevant Australian Standard, AS 2890.6, doesn't prescribe night-specific minimums, but the engineering decision on walking speed assumption should reflect actual users.
The third problem is detection. Inductive loop detectors work reliably at night, but video-based detection systems can lose accuracy in low-light conditions if camera settings or illumination haven't been validated for the night environment. A detector that fails to register a vehicle places the controller in a recall or rest-in-walk state that may not match actual demand. This is worth verifying at any intersection where adaptive control or actuation relies on video detection.
Designing a dedicated night-time timing plan
A purpose-built night-time plan starts with three engineering decisions: cycle length, pedestrian parameters, and phase split allocation.
Cycle length should be calculated from night-time volume data, not inferred from daytime counts. If the intersection has detection logs, review the 1 am–5 am window across at least 20 weekday nights and 10 weekend nights. Volume variation at night is higher in relative terms than during daytime peaks, so the plan needs to accommodate the 85th-percentile night volume rather than the average. This typically produces a cycle of 45–70 seconds for suburban arterial intersections, compared with 90–120 seconds during AM peak.
Pedestrian parameters should use a design walking speed of 1.0 m/s rather than the 1.2 m/s often used in daytime plans. This adds a few seconds to crossing time on wide roads and has negligible effect on vehicle throughput at low volumes. The clearance interval calculation doesn't change, but the minimum green for pedestrians does.
Phase split allocation at night can often simplify the phase sequence. If turning movements on a secondary approach drop below a threshold where a dedicated turn phase is warranted, combining those movements into a shared phase reduces lost time and keeps the cycle short. This is a controller configuration change, not a physical modification, and it can be scheduled as a time-of-day plan switch. The principle is the same one that governs adaptive signal control: when demand drops, the timing should follow it, not hold to a plan built for different conditions.
Time-of-day switching and plan transition
Controllers implement night-time plans through time-of-day scheduling, switching from the inter-peak plan at a defined clock time. The transition matters. Abrupt plan switches can cause a controller to jump mid-cycle into a new phase sequence, which creates a brief period of uncoordinated signal display if the intersection is part of a coordinated corridor. Best practice is to configure the switch to occur at an offset-neutral point in the cycle, or to use a transition period where the controller completes the current cycle before loading the new plan.
For coordinated corridors using green wave progression, night-time plans need matching offsets across the string of intersections. A plan that works in isolation can break coordination if the cycle length change shifts the offset relationship between adjacent sites. This isn't complex to configure, but it does require checking at the network level, not just the individual intersection.
Controllers running time-of-day switching should also log plan activation events. If a night-time plan fails to load due to a controller fault or clock drift, the intersection will default to whichever plan was active before. Without logging, that fault goes undetected until a review or complaint surfaces it.
Reviewing night-time plan performance
Night-time plan performance is rarely reviewed with the same frequency as peak plans. The metrics are harder to collect because floating observer counts at 2 am aren't practical, and most performance monitoring focuses on peak periods. Detection logs and controller event data are the practical alternative. Volume-per-phase, phase skip events, and pedestrian actuation counts all provide usable indicators of whether the night-time plan is serving actual demand.
A plan that shows consistent phase skipping (where the controller skips a phase because there's no demand) is signalling that the phase sequence is over-specified for night conditions. A plan where pedestrian actuation always maxes out the minimum green is signalling the opposite. Neither takes long to identify from a controller log export, and both point to specific configuration changes rather than a full re-timing exercise.
Night-time signal timing doesn't need to be complex. It needs to be deliberate. The conditions are different, the users are different, and the consequences of a poorly configured plan at 2 am are different from the same fault during the morning peak. Treating it as a distinct engineering problem, rather than a residual state, is the starting point.