ML Models Optimise Low-Cost Air Quality Sensor Network for Wider Deployment
The partners of Task 6.8 AI4IoT published a paper on machine learning techniques for low-cost IoT sensor calibration that improve the quality of data incoming from the network thus enhancing the deployment of air quality monitoring and decision-support of municipal smart systems.
The paper was published in a well-recognized journal with a high impact score the Sensors journal, volume 21, number 9. The paper is a result of the AI4IoT pilot on Air Quality Monitoring. This is one of the eight pilots in WP6 to be carried out within the AI4EU’s scope.
Using the data from an experiment in the city of Trondheim, Norway, the paper presents an approach towards automatic data calibration with the use of machine learning models. The ability of machine learning models to capture non-linear influences by external features (such as meteorological) shows to be efficient in the calibration procedure.
The paper illustrates examples of applications that are enhanced by calibrated data and fill user needs identified in talks with the local municipality, which go in the direction of new technologies for decision-making support for policymakers in the municipality and citizens.
The findings enable municipalities to utilise low-cost sensor networks while still having meaningful inputs to services that can be used by the municipalities themselves, and citizens in general, in decision making support. All experimenting services are to be deployed through the European AI on-demand platform, currently under development under the AI4EU project.
Full access to the paper is available at the link