Combining smartphone-embedded accelerometers and Artificial Intelligence toward increasingly accurate estimates of ground motion at local scale

Technological progress consisting of the development of low-cost miniaturized sensors smartphone embedded and 5G infrastructure makes an effective fostering of diffuse data acquisition and citizen science for the earthquake defence and seismic risk mitigation strategies.
We illustrate here the results of a project aimed at building an innovative solution based on Artificial Intelligence to provide increasingly accurate estimates of ground motion at local scale. The project has the aim of investigating the potential of enhancing the resolution of the earthquake impact reconnaissance
with respect to site effects. The first phase is based on a specifically developed platform that collects data by an App installed on users' smartphones and then analyses the seismic parameters (i.e. Peak Ground Acceleration, PGA and Peak Ground Velocity, PGV) received at the built-in mobile device accelerometers.
The test version of the Android mobile App uses a Machine Learning model trained on regional earthquake data.
3 pairs of Android smartphones placed on seismic stations located in the Umbria region was used to collect real data to be used for ML model improvement and
Proof of Concept finaliza􀆟on.