Sixteen months of adaptive operation, measured against the same LEDs.
A stretch that had already been relamped with LED and was already running on an astronomical clock. Adding full adaptive control took another 51 % off the bill — and the measurement went into a peer-reviewed journal rather than a brochure.
Road under adaptive control, with the weather station on the pole — installed site, Italy
Everything easy had already been done.
The luminaires were LED. The switching was astronomical, so nothing burned in daylight. The design class was correct for the road. In most projects that is where the story ends and the remaining consumption is treated as fixed.
It is not fixed. A main road is designed for the traffic it carries at its busiest, and it carries that traffic for a small part of the night. For the rest of it the standard itself allows a lower class — provided you can prove, continuously, that the conditions justify it.
LED luminaires · astronomical clock · fixed design class all night
The same luminaires · iLCS nodes · traffic, luminance and weather measured on the stretch · class recalculated minute by minute inside UNI 11248 / EN 13201
A node per luminaire, a few sensors per stretch.
The expensive part of adaptive lighting is not the hardware on the poles. It is having a network fast enough to carry a decision to every luminaire before the decision is stale.
Nodes
One per luminaire, driving the driver and measuring its own energy.
Traffic counting
Vehicles counted on the stretch every minute, with a ten-minute moving average behind the decision.
Luminance & weather
Measured road luminance closes the loop; fog suspends the reduction entirely.
Mesh & gateway
One cluster, no SIM per lamp, and a decision that survives a lost internet link.
Same road, same luminaires, declared baseline.
The comparison is against the same LED installation running on its astronomical clock — not against the discharge lamps that preceded it. That distinction is the whole difference between a measurement and a marketing number.
Sixteen consecutive months of normal operation — through summer nights and winter ones.
The energy measured by the nodes themselves, luminaire by luminaire, hour by hour, on the 0.8 km stretch (32 luminaires) covered by the study.
Pasolini et al., Sensors 24(18):5942, 2024 — with the University of Bologna and CNIT. Wi4B researchers are among the authors and the conflict of interest is declared in the paper.
All three from the same published study. The residential figure is lower because the design class is already low — there is less room above the minimum to give back.
How that compares with the usual alternatives.
Calculated in the same study on real traffic data, for a 100 W reference luminaire, by road class. These are computed figures, not parallel field measurements.
| Strategy | Class M3 | Class M4 | Class M5 |
|---|---|---|---|
| Full adaptive (measured traffic) | 56 % | 52 % | 22 % |
| Virtual midnight profile | 19 % | 18 % | 11 % |
| Motion sensing | 12–24 % | 5–11 % | 3–5 % |
The pattern is the point: on the roads that consume the most, a pre-programmed dimming profile recovers roughly a third of what measured adaptive operation recovers. On low-class residential roads the gap closes, and we say so before the tender does.
Beyond the energy.
Each node reports its own state, so a failed driver is a line in the morning list rather than a complaint.
Every reduction is recorded with the measurement that justified it — which is what an inspection asks for.
The network belongs to the operator: no SIM, no subscription attached to each luminaire.
Control node and sensor on a luminaire — references available on request
Have a road like this one?
Send us the length, the number of luminaires, the design class and a traffic count if you have one. We will come back with the class profile the standard allows and what it is worth per year.