Automatized calibration methods for low-cost air quality sensors

Artificial intelligence for wireless sensor networks


Within the project named FIRST (artiFicial Intelligence for wiReless Sensor neTworks) supported by the Stiftung Südtiroler Sparkasse of Bolzano through the Fusion Grant, Eurac Research is developing drift compensation and adaptive learning methods to extend the calibration lifespan of low-cost sensors in time. The project aims at comparing several models (multivariate linear regression, random searches, feedforward neural networks, etc.) by exploiting training-set and data-set from real sensors for air quality deployed in a urban area.

A main focus of FIRST is to push the computational workload toward the network’s edge. In this sense, the talk is connected to the SFScon’s topic Intelligent Sensors/Networks of the IoT panel. The contribution is relevant to SFScon cause the entire process will be automatized by means of open-source software components. Moreover, the training-set is consumed via the API of the Open Data Hub, which exposes air-quality data from the certified monitoring stations located in the province of Bolzano.