Project 57 · Smart Cities / Environment

Air Quality Monitoring Network

ML-Calibrated Low-Cost Sensors With Transparent Uncertainty

Industry: Smart Cities / Environment
Services: Sensor Design Calibration Cloud
TRL: 3 → 8
Duration: 7 months
Technologies: Electrochemical + optical sensors calibration ML public dashboards

Multi-sensor node + ML calibration vs reference stations — Figure 1 — Multi-sensor node calibrated by ML against reference monitoring stations.

Neighborhood AQI map with explicit uncertainty — Figure 2 — Neighborhood AQI map presenting estimates with explicit uncertainty bands.

Time-series validation + pre/post calibration scatter — Figure 3 — Time-series validation and pre/post calibration scatter against reference data.

Real-world Air Quality Monitoring Network deployment — Figure 4 — Real-world sensor node mounted on city infrastructure.

Project background

Cities want dense air-quality visibility that sparse reference-grade stations alone cannot provide. The client wanted a distributed network producing trustworthy data through calibration rather than expensive hardware.

Challenge

Lower-cost sensors drift and cross-interfere. Making their output meaningful required calibration against reference stations, environmental compensation, and ongoing validation.

Approach & solution

We designed a multi-sensor node with environmental compensation and an ML-driven calibration pipeline against reference stations. Public-facing dashboards show live air quality with clear uncertainty indicators, avoiding false precision.

Results & benefits

The network produced neighborhood-level air quality visibility aligned with reference stations within expected bounds. Public engagement with the dashboards was high during pilot periods.

Have a project in mind? Let's build it.

We reply within one business day.

Start a project