Multivariate Modeling to handle Urban Air Pollution Data observed trough Vehicular Sensor Networks
This work presents an interdisciplinary assessment that looks in-depth the tracking of air quality in urban environments. This kind of application is well suited to be approached with the paradigm of wireless sensor networks in their overall variations. Therefore a robust and diverse set of solutions have been developed following the technology capabilities advance.
The proposed experiment takes advantage of Vehicle Sensor Networks (VSN) by embedding sensor nodes to public transportation, addressing this study case with bus lines within such a way that the mobiles spread the sampling activity through a large number of different places visited during the route. Simultaneously, it alleviates restrictions of power management, packaging dimensions (size and weight), and overall maintenance issues. We perform environmental modeling based on real data considering a temporal and spatial multivariate behavior on observed phenomena. We define the city of São Paulo as reference for the case study and parsed the asserted data to create a multivariate map of samples, showing the behavior of five different air pollutants (CO, O3, PM10, NO2 and SO2) simultaneously while it also varies in time.
The current development stage covers a set of handling processes over an input data that has unformatted or missing information due to being sourced from real sensors and the creation of the map mentioned above.
Our methodology addresses
1) the mentioned environmental simulation,
2) deploying mobile sensor nodes and perform sensing process,
3) implement network activity and delivery of collected data,
4) predict a view of a monitored environment based on gathered data with the same strategy previously applied to create the initial field.
Finally, we plan to evaluate system-level performance and operational constraints through an event-based simulation, taking into account a detailed description of roads, bus lines, vehicle itineraries, and overall traffic information.
Environmental modeling, Vehicle Sensor Networks, Multivariate data analysis