The subject of this project is the development and implementation of innovative calculation methods in the area of artificial intelligence and machine learning, concerning the analysis, correction, processing, and forecasting of information on air pollution.

The data for analysis will be provided by means of a network of low-cost pollution sensors. A densely distributed network of such measuring devices (target 2-3 sensors/km^2 of the area covered by the measurement) solves three problems:

• data quality (thanks to continuous marking of the contamination level with data from many sensors that reduces the error of the measured value and enables data correction in case of damage to any sensor),
• immediate identification of local sources of contamination,
• collecting data from places not yet covered by the measurement (e. g. from areas of single-family buildings generating low emissions, and areas distant from precise but expensive measuring stations).

These three aspects are critical to the quality of forecasting and, for the first time, make it possible to assess the effectiveness of measures to protect air quality.