POLEBOX AI Siting Engine™ · the engine that picks the poles
POLEBOX hardware is deliberately simple. The value of the system sits in the model that decides which street lighting poles are worth equipping, and in the data that network produces every night, socket by socket.
The model
No guesswork. The AI Siting Engine measures how suitable a point in the city is for overnight charging, combining six variables drawn from public, free and independently verifiable sources. The unit of analysis is the ISTAT census tract, the finest statistical grid in Italy.
How many households in the zone have no garage or private parking space, and therefore no way to install a home wallbox.
The battery-electric and plug-in fleet actually present in the area, not national registration figures.
Where cars really park: bay geometry, parking rules, residential streets with long overnight dwell times.
Location and type of charging points already installed, to separate covered areas from uncovered ones.
The load the lighting circuit and the local grid can carry, the constraint that sets how many sockets fit on one line.
List, position and specification of the poles, the one input that is not open data: it comes from the municipality or the lighting operator.
Every source is public and citable. A score built this way stands up in front of a city council, an auditor and an investor, because anyone can redo the calculation.
Demonstration values, generated to illustrate the structure of the score. They do not refer to any real municipality or to an analysis that has been run.
The roadmap
One guiding principle: each phase commits resources only once the previous one has reduced the risk.
The unit of analysis is defined, the variables are selected, the weights are documented. This is the phase that makes the model reproducible: anyone using the same sources must reach the same result.
OutputThe scoring model specification, with documented variables and weights.
The model is run on a specific territory: every zone and every pole receives a 0 to 100 score, with a map and a ranking of the best poles. Development is AI-assisted, with startup speed and cost, and fully documented quality.
OutputThe Siting Dossier: the technical basis for deployment and a working document for the administration, repeatable in any new municipality.
With the network live, the central management system records every session, socket by socket and night by night, thanks to the MID fiscal meter and socket-specific NFC identification. Comparing predicted score against actual charging corrects the weights: the profile of a good pole is learned, and each new installation is more accurate than the last.
OutputA second-generation model validated on operating data. Accuracy no one can copy without installing first.
With thousands of live sockets, the same data supports demand forecasting, predictive maintenance and load management, meaning charging is concentrated when the grid is quiet and energy costs less. Generative AI enters as an application layer: automatic reading of municipal documents, dossier generation, natural language querying of the data.
OutputThe data platform that runs the network and accelerates expansion, city after city.
Method applied
DEA Solar Consulting has carried out an independent open-data siting study on an urban cluster in a major Italian city: 45 poles along roughly 1.4 km of residential corridor, including verification of supply points, voltage drop limits and cable sizing. It is a feasibility study on public sources, not a project approved by any administration.
Open standards
Model quality depends on measurement quality. Every upstream choice is therefore tied to open, verifiable standards.
Why it matters
Anyone can replicate the algorithm. No one can replicate every socket, every night: that data accrues only by installing.
This is why the first Italian pole-charging network is worth more than the second. AI as a barrier is worth ten times AI as a feature.
AI-assisted development, consistent with the Human + AI positioning of the project. POLEBOX Urban™ is among the startups selected for the Human + AI programme at SmartCityLab Milano.