You are here

Objectives - IMAQ

The aim of the IoT monitoring of air quality (IMAQ, PID2019-107910RB-I00, January 2020 - September 2023) project is to explore advanced data analysis techniques for improving the accuracy of the data acquired by low-cost IoT sensors for air monitoring. These techniques will be applied for developing system-wide techniques in IoT monitoring networks in more general settings. Our research is based on the use of experimental data obtained in real life conditions, and one of the outputs of our project is the publication as open data of datasets obtained in real life conditions and all the developed hardware and software.

We have investigated energy-saving techniques, based on duty-cycle and efficient use of storage and transmission to minimise energy consumption without using low-power communications. We have also chosen to use nodes designed specifically for the project instead of using commercial equipment. Thus, during the project we have developed the Captor version 4.0 (Captor4) nodes which include a Raspberry-PI as the main module, connected by an I2C bus to sensorisation sub-modules supporting temperature/humidity, electrochemical (NO2, O3 and NO) and particle (PM10, PM2.5) sensors.

Captor4 nodes have been deployed in 3 testbeds:

  • The first testbed, with 13 Captor4 nodes developed by our research group within the project, has been deployed in the city of Barcelona from May 2022 until the end of the project (September 2023), taking air quality measurements (NO2, O3, temperature and relative humidity). The nodes are still deployed after the project. These data have been exploited in research by members of the research and work team, and form the basis of studies that we continue to carry out in the research group.
  • A second test bed has been deployed, together with researchers from CSIC (IDAEA research group, Barcelona), at the reference station of Palau Reial (Barcelona), with 2 Captor4 nodes, with measurements of NO2, O3, temperature and relative humidity, and PMx (PM1.0, PM2.5 and PM10). The objective has been the measurement of black carbon. The data have been used in several joint research projects with CSIC-IDAEA.
  • A third test bed has been deployed on the UPC university campus, with temperature measurements on building surfaces, wind measurements (anemometers) and measurements based on thermographic cameras mounted on drones. This work is carried out in collaboration with New York University (NYU, USA), and the BSC-CNS.

We have used machine learning techniques to analyze data captured with IoT devices, mainly applied to air quality monitoring IoT networks. We have worked on advanced calibration techniques, and graph signal processing for analyzing data captured from sensor networks, to improve data quality, and to develop tools for automatic management of data quality as for example proxy sensors and virtual sensors. Some recent results are available at analysis of sensory data . For example, we have analyzed the calibration of sensors, including data fusion methods, using linear and non-linear machine learning models. We have also created a proxy for a black carbon sensor. We have also used graph signal processing techniques to reconstruct the signal in a sensor network, or to detect outliers using Volterra series.